Professional Documents
Culture Documents
The Monetary Transmission Process Recent Developme
The Monetary Transmission Process Recent Developme
Process
Recent Developments and Lessons for Europe
Edited by
The Deutsche Bundesbank
The Monetary Transmission Process
Edited by
ISBN 0–333–77244–X
Bundesbank.
332.4'94—dc21
00–042080
10 9 8 7 6 5 4 3 2 1
10 09 08 07 06 05 04 03 02 01
v
vi Contents
Discussion
Ignazio Angeloni 201
6 Differences Between Financial Systems in European Countries:
Consequences for EMU
Reinhard H. Schmidt 208
Discussion
Charles A. E. Goodhart 241
Comment
Alain Vienney 247
7 European Labour Markets and the Euro: How Much
Flexibility Do We Really Need?
Michael C. Burda 252
Comment
Hendrik J. Brouwer 276
Comment
LuõÂs Campos e Cunha 279
8 The Monetary Transmission Process: Concluding Remarks
Otmar Issing 283
Francesco Giavazzi 294
Claes Berg 298
Ignazio Visco 303
Manfred J. M. Neumann 311
Name Index 315
vii
Introduction
Heinz Herrmann
The debate on the monetary transmission process has not only continued in
the past few years but has also intensi®ed. There are a number of reasons for
this. In academic circles there are still various different approaches that seek to
explain how monetary policy in¯uences an economy. More and more areas
have come to be included in the analysis which attracted less attention in the
past. Thus, nowadays ± on the basis of an improved microeconomic theory ±
more attention is being paid to the role of ®nancial intermediaries and
®nancial systems in general than was previously the case.
In economic policy, central banks are confronted with the task of
safeguarding a level of price stability which, in most cases, is much greater
than that achieved in previous decades. How this target can best be achieved
is a question which depends in part on our state of knowledge of the
transmission process. In Europe, the debate has been given an additional
stimulus by the transition to European monetary union. How will this
changeover to a single currency in¯uence the effects of monetary policy in the
eleven member countries? Will the single monetary policy have similar
transmission effects in the different national economies or will marked
differences continue to exist? These are questions to which we are seeking to
®nd answers at the start of monetary union.
The Bundesbank therefore decided to invite economists from universities,
central banks and other institutions to a conference dedicated to these issues
on 26 and 27 March 1999. This volume contains the papers given and the
comments on them, with the introductory statements to a panel discussion
which concluded the conference (pp. 283±314).
Diverse models and methods have been developed in recent years aimed at
gaining a better understanding of the monetary transmission process. The
impression has arisen that the conclusions were not independent of the
methods of analysis used. For example, a study by the BIS (1995) found that,
depending on the type of model used, very different answers were provided to
the question of whether and how the transmission processes differ between
the countries examined.
1
2 Heinz Herrmann
problems: How is the general term `price stability' to be de®ned? How should
central banks formulate the rules which they follow? What are the lessons for
the European Central Bank (ECB)? Regarding the ®rst point, Svensson
discusses, inter alia, the advantages and disadvantages of in¯ation targets
versus price-level targets, the question of an appropriate price index for
formulating an in¯ation target and selecting the appropriate level for an
in¯ation target. Concerning the formulation of monetary policy rules,
Svensson draws attention to the advantages of forecast targeting over an
instrument rule and an intermediate targeting strategy. He then extends his
basic model to include non-linearities and stochastic coef®cients. He considers
a forecast targeting strategy which takes account of the probability distribu-
tion of the forecast to be an adequate response to such model extensions. He
advises the ECB to use a strategy of forecast targeting, to explicitly formulate
the in¯ation target, to publicise the in¯ation forecasts and to refrain from
assigning a special role to the money stock.
In his commentary, Mervyn King echoes Svensson's preferences for forecast
targeting, but warns against overemphasising the differences between the
strategies as applied in practice. He states it is more important that the
transmission mechanism is understood. He stresses the importance of looking
at the entire gamut of in¯ation forecasts when making monetary policy
decisions. Unlike Svensson, he considers it is justi®ed to assign a special role to
monetary growth in the monetary policy strategy. Jose Vin Ä als likewise sees
considerable advantages in a monetary policy strategy characterised as forecast
targeting by Svensson. His view is that such a strategy is consistent with major
principles of a good monetary policy practice, among which he includes
transparency and an orientation to the medium term. He expresses the view
that it is also much easier to communicate to the general public than an
optimal instrument rule.
Svensson has based his statements on monetary policy rules on a rather
generally formulated system of equations for describing the transmission
process in which a control variable, which is normally identical with a short-
term interest rate, stands in relation to predetermined variables (and possibly
forward-looking variables). In Chapter 3, `The Transmission Process', Allan
Meltzer looks more closely at this question, asking whether the importance of
money market rates or central bank rates is really that pivotal and exclusive in
the transmission process as is assumed in such models. Using examples from
US history, he seeks to show that focusing on a short-term interest rate is not
justi®ed. Rather, there had been periods in which monetary policy showed
effects which were not consistent with this usual conception of the monetary
transmission process. In his analysis he concentrates on historical phases of
de¯ation in which the development of the (real) monetary base explained
economic developments better than (real) short-term interest rates. Meltzer
mentions that two of these periods are suitable for showing that no liquidity
trap will emerge even when interest rates are low, a statement which takes on
4 Heinz Herrmann
added importance in the current phase of low interest rates. To support his
thesis of an effective wealth effect, he presents consumption functions in
which the real money stock has a clear explanatory value. He therefore
advocates keeping an eye on monetary growth when making monetary policy
decisions.
Empirical studies of the monetary transmission process which, on the one
hand, analyse the effects of the central banks' interest rate policy on the
economy and, on the other, ask which behaviour pattern is to be
recommended for monetary policy, are confronted with a severe methodo-
logical problem: it is not easy to decide whether the link between central bank
rates and/or money market rates and other economic variables such as
in¯ation is the result of a central bank's reaction function or whether it re¯ects
the implications of interest rate policy on the corresponding economic
variables. This prompts Axel Weber in Chapter 4, `Asymmetric Interest Rate
Policy in Europe: Causes and Consequences', to examine the link between
short-term interest rates and also in¯ation rates in the European countries (as
well as in the United States and Japan) and also the link between national
interest rates in Europe during the period of the European Monetary System
(EMS) and to demonstrate possible ways of better identifying the nature of the
links. Drawing on past papers (such as King and Watson, 1997), he uses
bivariate VAR models which take into account either interest rates and
in¯ation, or the German interest rate and that of another European country.
One of his ®ndings is that, in contrast to the United States, it is rather doubtful
whether German monetary policy can be accurately described by an interest
rate-smoothing rule if one assumes a long-run Fisher effect and long and
variable lags in monetary policy. Regarding the interest rate pattern in Europe,
he suggests that many central banks in Europe followed German interest rate
policy, but that there existed no simple long-run homogeneous pattern of
co-movement.
In his commentary, Carlo Favero agrees with the main thrust of Weber's
argument that considering monetary policy rules in isolation is a major
methodological problem. One should study such rules in the framework of
macromodels instead. However, he argues that bivariate models speci®ed in
®rst differences, as proposed by Weber, are only a ®rst step in the right
direction. It would be desirable to extend such a framework to achieve a closer
relation between theoretical and applied models. Philippe Moutot shares the
view that Weber raises an important question which should trigger a useful
research programme. Furthermore, he makes some more general remarks with
respect to monetary policy rules. He suggests that this chapter makes it clear
that it is dif®cult to identify the rules followed by the central banks under
consideration. Against this background, he warns against the hope that simple
rules can replace discretion. In his view the relevant issue is how to combine
the need for some limited discretion with the need for transparency vis-a-vis
the public, and describes how the ECB has proceeded in this respect.
Introduction 5
banks) is likely to have some advantages for the funding of innovative, young
enterprises. At the same time, he states that a comparison of the growth
performance of the United States and that of Germany does not indicate the
superiority of the US ®nancial system. Also, he takes the view that differences
between countries are a problem, especially whenever unexpected shocks
emanate from monetary policy. From this he concludes that monetary policy
should have a medium-term orientation as far as possible in order to avoid
such shocks.
References
BIS (1995) `Financial Structure and the Monetary Policy Transmission Mechanism',
Basle.
Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending,' in
N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press.
King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality', Federal Reserve Bank
of Richmond, Economic Quarterly, 83, pp. 69±101.
Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit
of Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT
Press.
Sims, C. A. (1998) `The Role of Interest Rate Policy in the Generation and Propagation of
Business Cycles: What has Changed Since the '30s?', in Federal Reserve Bank of Boston
(ed.), Beyond Shocks: What Causes Business Cycles?, Boston: Federal Reserve Bank of
Boston.
This page intentionally left blank
1
1 Introduction
11
12 Bennett T. McCallum
The ®rst of these is an IS-type relation representing demand for current output
± i.e. saving versus spending behaviour. Optimising theory suggests that the
variable Et yt1 should appear as indicated on the right-hand side; this has been
argued by Kerr and King (1996), McCallum and Nelson (1999), and Rotemberg
and Woodford (1997), among others. Neglect of this term would not simplify
derivations substantially and would not affect the main points of the
analysis.8 This future expected income term will accordingly be incorporated
in the quantitative models of Sections 3±4.
Equation (1.2) is one form of the P-bar price adjustment relation that was
rationalised and utilised in papers by McCallum and Nelson (1997, 1998). It is
not as widely used as the Calvo±Rotemberg (1983) model9 or variants of the
14 Bennett T. McCallum
and it is clear that 10 0, 11 . Also, since prices are fully predetermined,
22 23 0. Given these facts, the undetermined coef®cient procedure can
be used to ®nd the remaining values of the ij . The solution, it turns out, is
yt yt�1 1 =
1 1
1 � vt b1 1 =
1 1
1 � et
1:6
pt �
0 =1 �
b0 =b1 1
1 � =b1 1 yt�1
1:7
Now let us suppose that the above system were studied by some VAR approach
that correctly identi®ed the policy shock term et . The coef®cient
b1 1 /(1 1 )(1 ± ) is negative, so a VAR estimate of (1.6) would correctly ®nd
a negative effect from a positive shock to Rt , provided that there had been
enough sample-period variation in et . If et had stayed close to its mean value of
zero, however, statistical procedures might ®nd no signi®cant effect in ®nite
samples of a realistic size.
In the case of pt , (1.7) indicates that the policy shock et plays no role when
the effect of yt�1 is taken into account, as it would be in any VAR study. A study
that looked for monetary effects on in¯ation by the Granger-causality method
would therefore conclude that there are no such effects. The method relying
upon the `fraction of explained variance' of the various shocks would attribute
some portion to et since et�1 , et�2 , . . . help to explain yt�1 . But the fraction of
pt variability attributed to monetary shocks would then be precisely the same
as the fraction pertaining to yt variability. Note that since (1 ± )/b1 1 < 0,
Methodological Issues 15
these will be such that a surprise increase in Rt will have the effect of increasing
± rather than decreasing ± subsequent values of pt . This is a rather perverse
property of this simpli®ed model.12 But it is nevertheless true that an increased
value of the policy response parameter 1 will decrease the variability of
in¯ation, pt; in the system (1.1)±(1.3).
Furthermore, note that, since the unconditional mean Eyt equals zero by
construction, relation (1.1) implies that the average value of the real rate of
interest rt Rt ± Et pt1 equals ±b0 /b1 . Thus if the central bank chooses 0 to
equal ± b0 /b1 , as a sensible policymaker would do, (1.7) reduces to
Thus on average, over a large number of periods, realised in¯ation will tend to
equal , a target rate that is entirely determined by monetary policy.13 Within
a given policy regime, the `long-run' ± i.e. unconditional mean ± value of
in¯ation is entirely monetarily determined, but this fact would not be revealed
by shock-oriented VAR procedures.14
Having argued that it is more important to focus on the systematic portion
of monetary policy actions, rather than on shocks, we are then necessarily
driven toward the study of structural models, rather than VARs.15 The reason is
that even `identi®ed' or `semi-structural' VAR systems do not give rise to
behavioural equations that can be presumed to be structural ± i.e. policy-
invariant. The purpose of identi®ed VARs is to identify the unsystematic
component of monetary policy, not to generate policy-invariant equation
systems. But it is the latter that governs the effects of systematic or anticipated
policy actions. This should be emphasised, for it is the crucial point of my
argument. It is, I believe, consistent with the analysis and views of most
creators and practitioners of the identi®ed VAR approach, including Bernanke,
Blanchard and Watson, and Christiano, Eichenbaum, and Evans.16 My
position would not be accepted by Sims (1998b), however, so some discussion
of his approach is included below (in Section 6). It should be noted that my
objections to several aspects of VAR analysis are not the same as those put
forth by Rudebusch (1998). In fact, they are not actually objections to VAR
analysis per se but rather are arguments for concentrating on the systematic
component of policy rather than the shocks.
In this section the purpose is to describe one approach to analysis of the effects
of the systematic component of monetary policy. Since this undertaking
requires use of a structural model, according to the foregoing argument, the
results obtained will depend upon the adopted speci®cation of economic
behaviour. My starting point will be the small-scale, open economy, quarterly
model based on an optimising analysis that is developed and presented in
16 Bennett T. McCallum
McCallum and Nelson (1998). The following paragraphs will brie¯y outline
that model and report some simulation results that serve to characterise the
effects of monetary policy. Variants of the basic model will be considered in
Section 4.
Basing one's analysis on the assumption of explicit optimising behaviour by
the modelled individuals in a general equilibrium setting is obviously not
suf®cient ± and perhaps not necessary ± for the creation of a structural model
that is speci®ed with reasonable accuracy relative to economic reality. The
optimising general equilibrium approach can be helpful in this respect,
however, since it eliminates potential internal logical inconsistencies that are
possible when this source of intellectual discipline is absent. The model in
McCallum and Nelson (1998), henceforth termed the M±N model, has a very
simple basic structure since it depicts an economy in which all individuals are
in®nite-lived and alike. As with many recent models designed for policy
analysis, it assumes that goods prices are `sticky,' i.e. adjust only slowly in
response to changes in conditions. It differs from many previous efforts in this
genre, however, in three ways. First, the gradual price-adjustment speci®cation
satis®es the strict version of the natural-rate hypothesis.17 Second, the
modelled economy is open to international trade of goods and securities.
And, third, individuals' utility functions do not feature time-separability, but
instead depart in a manner that re¯ects `habit formation'.
This last feature is speci®ed as follows. A typical agent desires at t to
maximize Et
Ut Ut1 . . ., where the within-period measure Ut is speci®ed
as
Ut exp vt = � 1Ct =Ct�1 h �1= 1 � �1 Mt =pt 1� 1:8
Notes: The 2 values actually used are 1 /4 of the values listed; the latter correspond to units of
measurement in annualised percentage points, as are typically reported in the Taylor rule literature.
That statement applies to all results in this chapter. means MSV solutions are not available with
existing software.
Rt � Et�1 pt1 toward its long-run average value. As 1 is increased, with policy
response strength increased, the standard deviation of pt falls distinctly.
In the second panel, feedback response is taken from Et�1 pt1 � .
Clearly, the variability of pt is much greater than when Et�1 pt is the target
variable, especially for large values of 1 . In the model at hand, then, the
stabilising effect of monetary policy on the in¯ation rate is greater when
Et�1 pt , rather than Et�1 pt1 , is the variable responded to. That property
does not obtain for all model speci®cations, of course.
In both of the ®rst two panels, we see that application of stronger feedback to
in¯ation rate discrepancies has the effect of increasing the variability of y~ t , the
output gap. In the third panel we consider policy responses to the output gap, as
well as to in¯ation. In particular, we assume that the interest rate instrument is
adjusted upward when Et�1 y~ t is positive ± i.e. when output is expected to exceed
its natural rate value. As 2 is increased ± i.e. moving to the right in the table ± we
see that the variability of y~ t falls, as one would expect. Thus it is the case that
systematic monetary policy can be used to stabilise output (i.e. keep yt close to y t )
in this model, despite its highly classical long-run properties. When 2 is
increased with constant 1 and 3 , the variability of in¯ation increases.
Another way to see the effect of the systematic portion of monetary policy is
to compare impulse response functions for different values of policy rule
parameters. Let us return to the case represented in the ®rst panel of Table 1.1
i.e. with 2 0, 3 0.8, and 1 varied over a wide range. In this context, the
impulse response function for the policy shock itself is not as interesting as for
some of the other shocks. Let us consider ®rst a shock to the expectational IS
function ± i.e. a shock to preferences that increases the demand for current
Methodological Issues 19
4 Model speci®cation
The previous section has suggested some procedures for characterising the
effects of the systematic component of monetary policy for a given structural
model. But of course different models generate very different effects, so it is
essential to have a strategy for developing a good structural model. Most
researchers would agree that it is desirable for a model to be consistent with
both economic theory and empirical evidence, but that dual requirement is
only a starting point for consideration of numerous issues.
Like many economists, I have been persuaded that it is a desirable practice to
begin with the construction of a general equilibrium model in which
individual agents are depicted as solving dynamic optimisation problems. As
mentioned above, such a step is neither necessary nor suf®cient for obtaining
20 Bennett T. McCallum
0.2 0.2
0.1 0
y
q
0 –0.2
0 10 20 0 10 20
0 0
p
s
–0.5 –0.5
0 10 20 0 10 20
0 0.05
–0.1 0
R
Δp
–0.2 –0.05
0 10 20 0 10 20
0.2 0.4
0.1 0.2
q
y
0 0
0 10 20 0 10 20
0 0.5
0
S
p
–0.05 –0.05
0 10 20 0 10 20
0 0
Δp
–0.01 –0.02
R
–0.02 –0.04
0 10 20 0 10
20
b Responses to unit shock to IS; μ1 = 5.0
0.1 2
0.05 1
y
q
0
0
0 10 20
0 10 20
0
2
–0.5 0
p
s
–1 –2
0 10 20 0 10 20
0 0
–0.5 –0.05
Δp
–1 –0.1
0 10 20 0 10 20
0.4 4
0.1 2
q
y
0 0
0 10 20 0 10 20
0 5
–0.1 0
s
p
–0.2 –5
0 10 20 0 10 20
0.2 0
0 –0.1
R
Δp
–0.2 –0.2
0 10 20 0 10 20
0.4 1
0.2 0.5
q
y
0 0
0 10 20 0 10 20
0 2
–1 0
s
p
–2 –2
0 10 20 0 10 20
0 0.1
Δp
–0.1 0
R
–0.2 –0.1
0 10 20 0 10 20
– μ = 0.1
a Responses to unit shock to y; 1
0.4 1
0.2 0.5
q
y
0 0
0 10 20 0 10 20
0.05 1
0 0.5
p
–0.05 0
0 10 20 0 10 20
0.05 0.1
Δp
0 0
R
–0.05 –0.1
0 10 20 0 10 20
– μ = 5.0
b Responses to unit shock to y; 1
0 0
q
y
–0.5 –5
0 10 20 0 10 20
0 0
–1
s
p
–2 –5
0 10 20 0 10 20
1 1
0 0.5
R
Δp
–1 0
0 10 20 0 10 20
0 0
–1
–5
q
y
–2 –10
0 10 20 0 10 20
2 10
1 0
s
p
0 –10
0 10 20 0 10 20
0.4 2
0.2 1
R
Δp
0 0
0 10 20 0 10 20
0.4 0.4
0.1 0.2
q
y
0 0
0 10 20 0 10 20
0 0.2
–0.1 0
p
s
–0.2 –0.2
0 10 20 0 10 20
0 0
Δp
–0.05 –0.02
R
–0.1 –0.04
0 10 20 0 10 20
0.2 1
0.1 0.5
y
0 0
0 10 20 0 10 20
0 1
–0.5 0
p
–1 –1
0 10 20 0 10 20
0 0
Δp
–0.5 –0.1
R
–1 –0.2
0 10 20 0 10 20
b Responses to unit shock to IS; h = 0.0
Note:
a
Quarterly, 1955:1±1996:4. It should be noted that the the output gap in the ®rst column is
measured as in McCallum and Nelson (1997), not by any detrending procedure based only on the
output series itself.
With ! 0.5, this relation is almost identical to the F±M speci®cation, as has
been shown by Walsh (1998, pp. 224±5).27 Here ut re¯ects the random,
unsystematic component of pricing behaviour; it is assumed to be white noise.
For inclusion in our simulation analysis, numerical values must be attached to
and 2u E(u2t ). On the basis of results in Isard, Laxton, and Eliasson (1999),
I have adopted 0.0032 and 2u (0.0025)2 .28
The fourth panel of Table 1.2 reports RMSE values for simulations with the
same policy rule settings as before. As can be seen, the extent to which
in¯ation variability depends upon 1 (our measure of policy activism) is much
less than with the P-bar model. That is because the F±M speci®cation features
much more built-in in¯ation inertia, and also because of the presence in (1.10)
of the ut component that is not present in our P-bar cases. The sensitivity of y~ t
variability to the policy rule is, correspondingly, considerably greater than
with the P-bar model. Autocorrelation coef®cients for pt and y~ t are shown in
the two rightmost columns of Table 1.3. In keeping with our understanding of
the nature of the F±M speci®cation, in¯ation persistence is much greater than
with the P-bar model, and much closer to that found in the US data. The
persistence of y~ t is about the same as in our basic model ± i.e. quite substantial
although less than in the US data. In the ®fth panel of Table 1.2, the policy
28 Bennett T. McCallum
feedback rule responds to Et�1 pt1 rather than Et�1 pt , so as to determine
whether this type of `preemptive' response is more effective in the presence of
additional in¯ation inertia. As can be seen, however, the results are not greatly
affected by this change.
Impulse response functions for the F±M pricing speci®cation are shown in
Figures 1.6 and 1.7; the policy rule has 1 1.0 and responds to Et�1 pt1 . The
additional in¯ation inertia provided by this model shows up quite clearly in
the lower left-hand panels. It is worth noting that although our P-bar model
does not give rise to much in¯ation persistence, it does account for a
considerable amount of persistence in output.29 This ®nding con¯icts with
claims made recently by various writers, including Chari, Kehoe and
McGrattan (1995), Christiano, Eichenbaum and Evans (1997) and Andersen
(1998). The reason that such a disagreement is possible is that these authors all
presume that slow adjustment or staggering of goods prices is combined with
continuous clearing of the labour market. But what is assumed in the models
given above ± as well as in the work of McCallum and Nelson (1997,1998),
Taylor (1979, 1993b), Fuhrer and Moore (1995) and many others ± is that ®rms
produce whatever quantity is demanded at the prevailing price with labour
supplying as much labour as is needed (given capital, technology and the
production function). Current wages in this arrangement are irrelevant for
labour quantity determination, except via effects on prices, as in the
`instalment payment' discussion of Hall (1980). Labour supply conditions
are important only in the determination of y t , not yt ± y t .30 As a consequence,
these models do not imply that a contractionary monetary shock leads to a rise
in pro®ts, as suggested by Christiano, Eichenbaum and Evans (1997).
One important speci®cational issue that will not be explored quantitatively
concerns the absence of any monetary variables in the basic model of Section
3. In this respect that model is consistent with most recent analysis of
monetary policymaking, as represented in the NBER conference volume edited
by Taylor (1999). But is it actually reasonable to conduct monetary policy
analysis using an analytical framework that includes no money demand
function and indeed no reference to any monetary aggregate, either narrow or
broad? At a super®cial level, this question is answered by the well known point
that if a money demand function were appended to a basic model such as
(1.1)±(1.3), its only role would be to determine how much money would have
to be supplied to implement the interest instrument policy rule; implied paths
of yt , pt and Rt would be entirely unaffected.
At a less super®cial level, however, the question becomes one that asks
whether an optimising speci®cation, with the medium-of-exchange role of
money properly recognised, would yield an expectational IS function that
includes no real money balance terms. The answer to that question is that such
terms are absent only when the implied `indirect utility function' for the
optimising household is additively separable in consumption and real money
balances. Thus formulations of the expectational IS function of the type that
Methodological Issues 29
1 5
0 0
q
y
–1 –5
0 10 20 0 10 20
0 0
–0.5
s
p
–1 –5
0 10 20 0 10 20
0 1
Δp
–0.05 0
R
–0.1 –1
0 10 20 0 10 20
a Responses to unit shock to policy rule
0.2 0
0
q
y
–0.2 –0.05
0 10 20 0 10 20
0.05 0.05
0
p
0 –0.05
0 x10–3 10 20 0 10 20
5 0.01
Δp
0.005
R
0 0
0 10 20 0 10 20
b Responses to unit shock to IS
0.1 2
0 0
q
y
–0.1 –2
0 10 20 0 10 20
0.02 2
0.01 1
s
p
0 0
0 x10–3 10 20 0 x10–3 10 20
5 4
Δp
0 2
R
–5 0
0 10 20 0 10 20
a Responses to unit shock to UIP
1 2
0.5 1
y
0 0
0 10 20 0 10 20
0 2
–0.5 0
s
p
–1 –2
0 10 20 0 10 20
0 0.1
Δp
–0.05 0
R
–0.1 –0.1
0 10 20 0 10 20
bb Responses
Responses toto unit
unit shocktotoytY
shock
–
are prevalent in the literature ± and used above ± depend upon this separability
assumption. To evaluate whether such an assumption is appropriate, one
would have to consider alternative ways of modelling the role of the medium
of exchange ± e.g. money in utility function, shopping time, transaction cost
or cash-in-advance setups ± and alternative functional forms (complete with
quantitative properties). Such a study is far beyond the scope of the present
paper, so I will end this discussion simply by noting that separability is not
compatible with the shopping-time formulation utilised by McCallum and
Goodfriend (1987).
5 Model diagnostics
p y~ R
Notes:
a
Variance statistics are quarterly fractional units multiplied by 104 in all panels.
b
Standard deviations are in percentages, annualised for p and R, in all panels.
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
–
Autocorrelation
Autocorrelationfunctions forΔΔp,yỹ,, RR in
functionsfor in US
US data
date
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
p, ỹy~,, R
Δp,
Autocorrelation functions for Δ R ininbasic
basicmodel
model
the autocorrelation magnitudes and patterns re¯ect the nature of the dynamic
interrelationships. Any major discrepancy on any of these dimensions ± any
discrepancy between a model's properties and actual data ± re¯ects a weakness
in the model's speci®cation. This argument presumes, clearly, that the policy
rule in the model simulations and the various shock variances are realistically
matched to the ones that prevailed over the sample period during which the
data was generated.
The foregoing objection to impulse response methods does not pertain, of
course, to VAR systems in which all shocks, not just the one associated with
monetary policy actions, are identi®ed. Examples are provided by Sims
(1998a), Blanchard and Watson (1986) and many others. But the identifying
restrictions in these systems are much more demanding and less credible than
in the semi-structural systems promoted by Bernanke and Mihov (1998a),
Christiano, Eichenbaum and Evans (1998) and others who seek robustness.
But whether this point of view is persuasive or not, the vector autocorrelation
strategy seems at least somewhat attractive because of its purely descriptive
nature (as mentioned by Fuhrer and Moore, 1995). Accordingly, an illustrative
34 Bennett T. McCallum
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
same sign in the model and in the data, but only one (for p and R) represents
a reasonably close quantitative match.
The third panel in Table 1.4 and Figure 1.10 pertain to the same model
except with h 0 ± i.e. with habit formation eliminated from the households'
saving decision. Surprisingly, this elimination slightly increases the persis-
tence of in¯ation. But it does not overcome the other major problems with the
basic model.
Next we turn to the model in which the price adjustment (1.10) replaces the
P-bar speci®cation. Now the variance magnitudes, reported in panel d of
Table 1.4, are closer to those in the data. And the own autocorrelation
functions shown in Figure 1.11 are distinctly more similar to those of Figure 1.8.
Indeed, they provide a match that might be judged as semi-respectable. But
the cross autocorrelations match quite poorly, especially those involving y~ .
Thus this chapter's ®ndings are basically consistent with those reported by
Fuhrer (1997, 1998).33 Setting h 0, in Figure 1.12 and panel e of Table 1.4,
worsens the match between model and data, especially in terms of the
variance magnitudes.
36 Bennett T. McCallum
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
Autocorrelation functions for Δp, y˜y~,, RR in
Δp, in model
model with
with (1.10) and hh == 0.0
(1.10) and 0.0
6 Other approaches
Considerable attention has been devoted, during recent years, to the `narrative
approach' to measuring the effects of monetary policy that was pioneered by
Romer and Romer (RR) (1989). As is well known, the RR study generated a
Methodological Issues 37
base was strongly in¯uenced by a measure of current bank failures (prior to the
creation of the FDIC).34
Another VAR-oriented approach to the study of systematic monetary policy
responses was developed by Bernanke, Gertler and Watson (1997), who
concluded that a large fraction of the US economy's real effects from oil price
shocks since 1970 has resulted from the monetary policy response to these
shocks, rather than from the shocks themselves. The study is concerned with
attributing historical ¯uctuations to various sources, rather than with the type
of characterisation attempted in the present chapter. Bernanke, Gertler and
Watson (1997, p. 93) state that their method `certainly is not invulnerable to
the Lucas critique'.
Closer in spirit and approach to the present chapter is a study by Dotsey
(1999). It, too, utilises simulations with a setup that features maximising
behaviour and aspires to the development of a policy-invariant, structural
model, and it reaches similar conclusions. One major difference relative to the
present chapter is that Dotsey's comparisons are made across entirely different
policy rule speci®cations, rather than across different parameter settings for
variants of a single rule, as is typically the case in Sections 3±5 above.
7 Conclusions
Let us conclude with a very brief summary. The chapter has argued that, in
studying the monetary policy transmission process, more emphasis should
be given to the systematic portion of policy behaviour and correspondingly
less to random shocks ± basically because shocks account for a very small
fraction of policy instrument variability. Analysis of the effects of the
systematic part of policy requires structural modelling, rather than VAR
procedures, because the latter do not give rise to behavioural relationships
that can plausibly be regarded as policy-invariant. By use of an illustrative
structural speci®cation with variants, the chapter characterises the effects of
policy parameter settings by means of impulse response functions and root-
mean-square statistics for target errors. Different models give different
answers to questions about the effects of systematic policy, so procedures for
scrutinising model speci®cation are essential. In this regard, it is argued that
vector autocorrelation functions, augmented by variance statistics for each of
a model's variables, seem more promising than impulse response functions
because the latter require shock identi®cation, which is inherently a dif®cult
process.
Notes
1. I am indebted to Miguel Casares, Larry Christiano, Marvin Goodfriend, Jeffrey
Lacker, Ellen McGrattan, Allan Meltzer, Edward Nelson and Harald Uhlig for
helpful commments and suggestions.
Methodological Issues 39
23. This difference has been usefully emphasised by Fuhrer (1997, 1998). A paper by
Estrella and Fuhrer (1998) stresses that `standard models' with h 0 and a Calvo±
Rotemberg price adjustment speci®cation are seriously inconsistent with the data. I
agree with that judgement but see no reason to conclude that all optimising models
with rational expectations are unsatisfactory in this regard.
24. To close down the basic model, it is necessary not only to eliminate exports and
imports, but also to adjust the variance of the shock term driving y t (because the
latter no longer depends on imported raw materials). For more explanation, see
McCallum and Nelson (1998).
25. To re¯ect price level stickiness, some departure from full optimizing behaviour is
required ± e.g. some additional constraint must be imposed relative to a ¯exible
price general equilibrium system. There is considerable scope for dispute
concerning the relative `rationality' of different departures.
26. The Fuhrer±Moore (1995) paper does not use optimising analysis to generate its
consumption behaviour, but instead posits a non-expectational IS function. Also it
uses detrended yt as its measure of the output gap.
27. The difference is that (1.10) includes only y~ t in place of 0.5Et (y~ t y~ t 1 ).
28. To get 0.0032 from the Isard, Laxton and Eliasson (1999) value of 3.2, one divides
by 400, because they express in¯ation in annualised percentage points, and then
divides by 2.5 to re¯ect the slope of an Okun's Law relationship between y~ t and
unemployment.
29. Simulation results indicate that yt features signi®cantly more persistence than does y~ t .
30. Recall that y t is the natural rate value of yt that would prevail in the absence of any
nominal frictions.
31. There exists some controversy even over the robustness of these procedures. For
recent contributions on this topic, see Faust (1998) and Uhlig (1999).
32. It would be possible to judge a model's ®t entirely on the basis of the impulse
response functions for the policy shock, as in Rotemberg and Woodford (1997), but
this seems undesirable given the small contribution to overall variability coming
from this source.
33. They are also somewhat in the spirit of Estrella and Fuhrer (1998), but do not
involve the Calvo±Rotemberg pricing equation that was a component of the MN
model criticised by Estrella and Fuhrer.
34. My study found that a policy rule that adjusts the monetary base so as to attempt to
keep nominal income on a steady growth path would have made the 1930s' fall in
nominal income much less severe than actually occurred. This activist feedback
rule would have resulted in a much greater expansion of M1 than in Christiano's
counterfactual simulation (which was in turn more expansionary than Sims').
What this monetary stimulus would have done for real output depends, of course,
on the model utilised.
References
Andersen, T. M. (1998) `Persistency in Sticky-price Models', European Economic Review,
42, p. 593±603.
Bernanke, B. S. (1986) `Alternative Explanations of the Money-income Correlation',
Carnegie±Rochester Conference Series on Public Policy, 25, 49±99.
Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of
Monetary Transmission', American Economic Review, 82, 901±21.
Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of
Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48.
Methodological Issues 41
Bernanke, B. S., M. Gertler and M. W. Watson (1997) `Systematic Monetary Policy and
the Effects of Oil Price Shocks', Brookings Papers on Economic Activity, No. 1, 91±142.
Bernanke, B. S. and I. Mihov (1998) `Measuring Monetary Policy', Quarterly Journal of
Economics, 113, 869±902.
Blanchard, O. J. and M. W. Watson (1986) `Are Business Cycles All Alike?', in R. J.
Gordon (ed.), The American Business Cycle, Chicago, University of Chicago Press.
Calvo, G. A. (1983) `Staggered prices in a Utility-maximizing Framework', Journal of
Monetary Economics, 12, 383±98.
Chari, V. V., P. J. Kehoe and E. R. McGrattan (1996) `Sticky Price Models of the Business
Cycle: Can the Contract Multiplier Solve the Persistence Problem', NBER Working
Paper 5809.
Christiano, L. J. (1999) `Discussion', in Beyond Shocks: What Causes Business Cycles?,
Boston, Federal Reserve Bank of Boston.
Christiano, L. J. and M. Eichenbaum (1992) `Identi®cation and the Liquidity Effect of a
Monetary Policy Shock', in A. Cukierman, Z. Hercowitz and L. Leiderman (eds) Political
Economy, Growth, and Business Cycles, Cambridge, MA, MIT Press.
Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited
Participation Models: A Comparison', European Economic Review, 41, 1201±49.
ÐÐÐÐ (1998) `Monetary Policy Shocks: What Have We Learned and to What End?',
NBER Working Paper 6400.
Clarida, R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice: Some
International Evidence', European Economic Review, 42, 1033±68.
Cochrane, J. H. (1994) `Shocks', Carnegie±Rochester Conference Series on Public Policy, 41,
295±364.
Dotsey, M. (1999) `The Importance of Systematic Monetary Policy for Economic
Activity', Federal Reserve Bank of Richmond Economic Quarterly, 85, 41±59.
Dotsey, M. and P. N. Ireland (1995) `Liquidity Effects and Transaction Technologies',
Journal of Money, Credit, and Banking, 27, 1441±56.
Estrella, A. and J. C. Fuhrer (1998) `Dynamic Inconsistencies: Counterfactual Implica-
tions of a Class of Rational Expectations Model', Federal Reserve Bank of Boston,
Working Paper 98±5.
Faust, J. (1998) `The Robustness of Identi®ed VAR Conclusions about Money', Carnegie±
Rochester Conference Series on Public Policy, 49, 207±44.
Fuhrer, J. C. (1997) `The (UN)importance of Forward-looking Behavior in Price
Speci®cations', Journal of Money, Credit, and Banking, 29, 338±50.
ÐÐÐÐ (1998) `An Optimization-based Model for Monetary Policy Analysis: Can Habit
Formation Help?', Boston, Federal Reserve Bank of Boston, Working Paper.
Fuhrer, J. C. and G. R. Moore (1995) `In¯ation Persistence', Quarterly Journal of Economics,
110, 127±59.
Hall, R. E. (1980) `Employment ¯uctuations and Wage Rigidity', Brookings Papers on
Economic Activity, 1, 91±123.
Isard, P., D. Laxton and A.-C. Eliasson (1999) `Simple Monetary Policy Rules under
Model Uncertainty', IMF Working Paper 99/75; also in P. Isard, A. Razin and A. K. Rose
(eds) International Finance and Financial Crises, Boston, Kluwer Academic Publishers.
Kerr, W. and R. G. King (1996) `Limits on Interest Rates Rules in the IS Model', Federal
Reserve Bank of Richmond Economic Quarterly, 82, 47±75.
Klein, P. (1997) `Using the Generalized Schur Form to Solve a System of Linear
Expectational Difference Equations', University of Stockholm, Working Paper.
Leeper, E. M. (1997) `Narrative and VAR Approaches to Monetary Policy: Common
Identi®cation Problems', Journal of Monetary Economics, 40, 641±58.
42 Bennett T. McCallum
Leeper, E. M., C. A. Sims and T. Zha (1996) `What Does Monetary Policy Do?', Brookings
Papers on Economic Activity, 2, 1±63.
Lucas, R. E., Jr. (1972) `Econometric Testing of the Natural Rate Hypothesis', in
O. Eckstein (ed.), Econometrics of Price Determination, Washington, DC, Board of
Governors of the Federal Reserve System.
McCallum, B. T. (1990) `Could a Monetary Base Rule have Prevented the Great
Depression?', Journal of Monetary Economics, 26, 3±26.
ÐÐÐÐ (1999) `Role of the Minimal State Variable Criterion in Rational Expectations
Models', Working Paper, NBER Working Paper 7087; also in P. Isard, A. Razin and A. K.
Rose (eds) International Finance and Financial Crises, Boston, Kluwer Academic
Publishers.
McCallum, B. T. and M. S. Goodfriend (1987) `Demand for Money: Theoretical Studies',
in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of
Economics, Macmillan, also in J. Eatwell, M. Milgate, P. Newman, The New Palgrave
Dictionary of Money and Finance, London.
McCallum, B. T. and E. Nelson (1997) `Performance of Operational Policy Rules in an
Estimated Semi-classical Structural Model', NBER Working Paper 6599; also in Taylor
(1999).
ÐÐÐÐ (1998) `Nominal Income Targeting in an Open-Economy Optimizing Model',
NBER Working Paper, 6675, also in Journal of Monetary Economics, 43, 553±78.
ÐÐÐÐ (1999) `An Optimizing IS-LM Speci®cation for Monetary Policy and Business
Cycle Analysis', Journal of Money, Credit, and Banking, 31, 296±316.
Meltzer, A. H. (1995) `Monetary, Credit (and Other) Transmission Processes: A
Monetarist Perspective', Journal of Economic Perspectives, 9, 49±72.
Mishkin, F. S. (1995) `Symposium on the Monetary Transmission Mechanism', Journal of
Economic Perspectives, 9, 3±10.
Nelson, E. (1998) `Sluggish In¯ation and Optimizing Models of the Business Cycle',
Journal of Monetary Economics, 42, 303±22.
Obstfeld, M. and K. Rogoff (1995) `The Mirage of Fixed Exchange Rates', Journal of
Economic Perspectives, 9, 73±96.
Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit of
Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT Press.
Rotemberg, J. J. (1982) `Monopolistic Price Adjustment and Aggregate Output', Review of
Economic Studies, 44, 517±31.
Rotemberg, J. J. and M. Woodford (1997) `An Optimization Based Econometric
Framework for the Evaluation of Monetary Policy', NBER Macroeconomics Annual
1997, Cambridge, MA, MIT Press.
ÐÐÐÐ (1999) `Interest Rate Rules in an Estimated Sticky Price Model', NBER Working
Paper 6618.
Rudebusch, G. D. (1998) `Do Measures of Monetary Policy in a VAR Make Sense?'
International Economic Review, 39.
Shapiro, M. D. (1994) `Federal Reserve Policy: Cause and Effect', in N. G. Mankiw (ed.),
Monetary Policy, Chicago, University of Chicago Press.
Sims, C. A. (1992) `Interpreting the Macroeconomic Time Series Facts: The Effects of
Monetary Policy', European Economic Review, 36, 975±1000.
ÐÐÐÐ (1998a) `Comment on Glenn Rudebusch's ``Do measures of monetary policy in
a VAR Make Sense?' '', International Economic Review, 39.
ÐÐÐÐ (1998b) `The Role of Interest Rate Policy in the Generation and Propagation of
Business Cycles: What has Changed Since the 30's?', in Federal Reserve Bank of Boston,
Beyond Shocks: What Causes Business Cycles?, Boston, Federal Reserve Bank of Boston.
Methodological Issues 43
Lawrence J. Christiano
1 Introduction
The perspective that I have on monetary policy shocks differs from that of
McCallum, and I begin my discussion by explaining why. In my discussion of
the use of second moments in empirical analysis, I relate McCallum's analysis
to the existing literature. Finally, in proposing the use of a historical episode to
help evaluate a model, I seek to add another empirical strategy to the set
considered by McCallum. I argue that evaluating a model's ability to account
for a particular episode can be a useful complement to the overall assessment
of a model. The episode I consider is the take-off of in¯ation in the early 1970s,
an experience that I believe represents an important challenge to McCallum's
sticky-price model, and other sticky-price models as well.
44
Discussion 45
St f
t "t
2:1
Here, St denotes the variable (or, vector of variables) that the central bank
controls directly. McCallum follows much of the recent literature by treating St
as a short-term interest rate. The central bank's decision variable is assumed to
be determined in part as a systematic function, f ; of the state of the economy,
One way to test a model, not considered by McCallum, is to see how well it
accounts for the events of a particular historical episode. Given the motivation
for the type of work in this chapter, I believe there is one episode that deserves
special attention. As I understand it, a primary motivation for research on
monetary policy rules is to identify strategies for conducting monetary policy
that will ensure that we not repeat the monetary policy mistakes of the past.
One such mistake is the burst of in¯ation experienced by many countries in
the 1970s. In view of this, a natural test of a model like McCallum's is to see
whether it can account for this episode. After all, if the cures for past monetary
disorders emerging from a particular model framework are to be compelling,
48 Lawrence J. Christiano
f
t constant 3 Rt�1
1 � 3
1 1 Et t1 2 y t
where t1 is the in¯ation rate in the quarter following quarter t; Et is the
conditional expectation operator and yt is the deviation of output from
potential. A feature of this model is that when 1 < 0; then it is possible for
higher anticipated in¯ation to be self-ful®lling. Since nothing in this result
depends on the value of 3 ; to simplify the intuition I set 3 0: Under these
circumstances, higher expected in¯ation can be self-ful®lling via the following
mechanism.5 Suppose agents expect higher in¯ation. With 1 < 0; this creates
the expectation that the real interest rate will be low, which in turn stimulates
investment demand. The rise in investment demand then triggers a rise in
output and employment which ultimately translates into higher in¯ation,
thereby validating the original jump in in¯ation expectations.
Note that, according to McCallum's model, a burst of high in¯ation
triggered by expectations in this way is associated with a strong economy. A
problem with this argument, however, is that variables like investment and
employment were actually low in the early 1970s, particularly during the 1975
recession.
An alternative model that has been used for monetary analysis is the limited
participation model studied by Christiano (1991), Fuerst (1992) and
Christiano, Eichenbaum and Evans (1998b). Christiano and Gust (1999,
2000) show that this model also has equilibria in which higher expected
in¯ation can be self-ful®lling when 1 < 0: However, Christiano and Gust
(1999b) show that in these equilibria, output and investment are low, a feature
that appears more consistent with the data.
Thus, it appears that the sticky-price model of McCallum fails the test of the
1970s, a test that the limited participation model passes. There is, however,
one important caveat that must be mentioned in this context. There are many
things that happened in the 1970s ± the productivity slowdown began, there
were oil shocks ± and possibly some of the evidence of weakness that
Discussion 49
5 Conclusion
Notes
1. This section summarises arguments in Christiano, Eichenbaum and Evans (1998a).
2. For a discussion of this model, see Cooley and Hansen (1997).
3. See Fuerst (1992); Christiano (1991); Christiano, Eichenbaum and Evans (1998) and
Alvarez, Atkeson and Kehoe (1999).
4. For an informal discussion of how this might have happened, see Blinder (1982),
especially p. 264. Formal analyses appear in Chari, Christiano and Eichenbaum
(1998) and Clarida, Gali and Gertler (1998).
5. The observation that 1 < 0 can result in equilibria in which in¯ation expectations
are self-ful®lling in a simpli®ed version of McCallum's model (an IS±LM model with
rational expectations) was ®rst made in Kerr and King (1996). Ben McCallum was
kind enough, in a private communication, to con®rm that this result also holds in
the models analysed in his chapter.
6. In Christiano and Gust (2000)'s analysis, they simulate Clarida, Gali and Gertler's
(1998) model and a limited participation model under the assumption that the jump
in in¯ation expectations in the early 1970s was a reaction to a bad technology
shock. This is consistent with the view that the higher in¯ation expectations at the
time were in part due to the rise in oil and other commodity prices that occured in
the wake of the shortages that occurred at the time. The policy rule used in the
simulations uses the parameter values estimated by Clarida, Gali and Gertler (1998)
using data from the 1960s and 1970s.
References
Alvarez, F., A. Atkeson and P. Kehoe (1999) `Money and Interest Rates with
Endogenously Segmented Markets', NBER Working Paper, 7060.
50 Lawrence J. Christiano
Harald Uhlig
1 Introduction
51
52 Harald Uhlig
public, for framing the debate or for providing a ®rst, `rough' benchmark of
comparison. Opponents like to use them to caricature the academic point of
view. But these rules are really meant as a starting point for the analysis and its
debate. Chapter 1 by Ben McCallum is a insightful and pathbreaking
contribution to that end.
The chapter provides an outline of a methodology for the analysis of
monetary transmission mechanisms. Together with a few other leading
researchers, McCallum aims at applying modern advances in macroeconomic
theory to practical issues of monetary policy. One can brie¯y summarise that
methodology as follows. First, write down one or better several models of the
economy with solid microfoundations. Second, show with a careful
comparison with the data, that the model indeed captures the observed
behaviour of the economy. Third, experiment with suggestions for policy
changes, using the models at hand. Fourth, provide a detailed, insightful and
understandable account of the effects of policy and their changes to the policy
maker. Ultimately, the aim must be to provide the policy maker with a deeper
intuition about the cause and the effects of monetary policy and their changes.
Rather than discuss the methodology in the abstract, McCallum provides a
speci®c application, studying the Taylor rule in the context of a model
framework developed by him and Nelson in a number of previous
contributions.
I have little doubt that McCallum's pioneering and ambitious contribution
points us in the right direction. I liked the chapter a lot. If there are any
disagreements, then they are with the details of the arguments and with the
quality of the theoretical advances achieved so far. As a discussant, my role is
to emphasise these points of departure rather than to restate the compara-
tively larger parts where author and discussant agree. In summary: the
approach is a complement not a substitute to other approaches, in particular
the VAR-based methodology, and the theory needs further development.
Theories with solid microfoundations, on the other hand, are more suitable
to contemplate changes in the systematic component f
. One may wonder,
whether the systematic component has any effect at all other than in¯uencing
the rate of in¯ation. One may also wonder whether there truly are changes in
the systematic component: perhaps, the continuous trial-and-error evolution
of the way monetary policy responds to current circumstances should more
appropriately be understood to be part of the shock component t . But be this
as it may: one certainly has to conclude that VARs and economic theory are
complements, not competitors. Different problems require different tools.
Good, advanced theories of money are still remarkably hard to come by. There
is a simple reason: this is a very tough objective! Further advances here are not
only desirable, they are crucial.
3 The Model
The results in the chapter are based on an analysis of the model in McCallum
and Nelson (1998). Despite the contrary appearance in chapter 1, the model is
constructed according to the current standards in modern macroeconomic
theory ± i.e. is of the stochastic, dynamic, quantitative, rational expectations
applied dynamic general equilibrium variety. I, for one, ®nd it unfortunate,
that the deep beauty and insight of these models apparently still has to be
`sold' using the now outdated and cruder IS±LM language, but perhaps this is
still necessary for another decade or two to reach the broadest audience
possible. McCallum has rightly shown and emphasised that the IS±LM logic is
not in con¯ict with these models, and, indeed, it isn't. The inherent logic
provides an improvement on the IS±LM view, however. To take a perhaps
strained analogy: there also is no con¯ict between the view that most animals
are near-perfect at the tasks they do as if created by divine intervention for that
purpose, and the view that this near-perfection is the result of genetic,
Darwinian evolution: the latter view simply is more helpful for understanding
what we see and for predicting what will happen if the environment changes.
Likewise, a well built model in the modern tradition is going to be reliable and
more helpful for analysing policy changes than the outdated and inferior IS±
LM framework.
The model by McCallum has a number of special features. The economy is
modelled as open. Investment is exogenous. Labour supply and thus output
gaps are demand-driven. Firms may ®nd themselves producing below
capacity. There is habit formation in consumption.
4 Exogenous investment?
the most interesting one for monetary policy. Whether directly through
interest rates or indirectly through the credit channel, monetary policy mainly
affects investment.
The effect on consumption via a `price channel' is much too sluggish to be of
major importance. Of course, having a more explicit role for investment can
be done in the context of this model. It makes the model less elegant, but not
much more complicated, but it would then have an additional and appealing
channel for monetary policy. A key dif®culty here would be to reconcile the
observed ¯uctuations in interest rates with the observed ¯uctuations in
investment, a point not suf®ciently emphasised by the current real business
cycle literature.
It is too easy to write an elegant theoretical model, and too dif®cult to write
a model that also replicates key dynamic elements found in the data.
Policymakers will rightly be leery of the former, and at least somewhat more
comfortable with the latter.
It would be hard to disagree with this statement. How well does McCallum's
model do? McCallum's Chapter 1 invites the reader to investigate this issue. It
provides a number of comparisons to the data, that are similar in spirit to the
practice in the real business cycle literature of comparing (®ltered) correlations
at leads and lags, although more could be provided: Figure 1.8 (p. 32) offers
only a partial assessment. There is also no principal difference between
comparing raw correlation patterns versus comparing impulse response
functions in (reduced-form) VARs: both VARs and correlation patterns provide
summary information about the stochastic properties of the models and the
data-and, in fact, one can essentially compute the former from the latter and
vice versa. Which representation one ®nds more useful thus seems to be purely
a question of taste, and I would have found it helpful to also provide a
systematic comparison of the latter.
The comparison to the data in Tables 1.3, 1.4 and Figures 1.8±1.12 (pp. 27±
36) shows that the differences to the data are still uncomfortably large. Policy
makers should still be leery of the model at hand. That said, one should
understand how dif®cult an exercise this is. While we are not at the end of the
road yet, the chapter nonetheless provides a big step forward. Much to his
credit, McCallum discusses all these issues openly and frankly rather than
hiding them for someone else to ®nd. This is an honest, scienti®c approach
which should serve as the model for how things should be done.
Table 1.4 shows particularly big differences with the data regarding the
in¯ation variability, except for model d. Should the reader conclude that one
Discussion 55
e �
R R � e
R � R
56 Harald Uhlig
For monetary policy and its effect on the economy, is there a sizeable
difference between reacting to current in¯ation Et�1 pt versus reacting to
in¯ation forecasts, Et�1 pt1 ? It is here where the chapter makes a direct
contribution to an intense and current policy debate of direct practical
consequence: the answer sheds important light on the question, whether
monetary policy should be targeting in¯ation forecasts or whether it should
react to current developments in prices.
Table 1.1 seems to provide the answer. Indeed, McCallum writes, that the
stabilising effect of monetary policy on the in¯ation rate is greater when
Et�1 pt , rather than Et�1 pt1 , is the variable responded to. Obviously, this
is true in the sense of tracing out the results of changing the numerical value in
the policy rule. But is it true in the relevant economic sense? Figures D1.1 and
3.5
3
2.5
E(t–1
t –(Δ1p) t)Infl(
ast) as
Output variability
2 target
1.5 t –(Δ1p) t+1
E(t–1 ) t + 1)
Infl(
1 as target
0.5
0
0 5 10 15
Inflation variability
Figure D1.1 Comparison of two rules: the `Taylor menu': in¯ation and output
Figure D1.1 shows the `Taylor menu', juxtaposing the variance of in¯ation with the
variance in output, using the numbers from the ®rst two rows in Table 1.1 in the
chapter. As one can see, the results form a common line.
Discussion 57
4
Interest rate variability
t –(Δ1p) t)Infl(
E(t–1 ast) as
3 target
2 t –(Δ1p) t+1
E(t–1 ) t + 1)
Infl(
as target
0 5 10 15
Inflation variability
8 The future
Central banks need to analyse the effects of policy changes. Insights into the
effects of changes to the systematic `feedback' part of policy are particularly
important: how should monetary policy react to developments in the
economy? Providing substantive answers requires carefully spelled out
theories, in which potential policy changes can be examined. These theories
should make use of the tools of modern macroeconomics. They should be
stochastic, dynamic, quantitative, rational expectations applied general
equilibrium models. McCallum is one of the leading pioneers of that new
direction, paving the way. These theories are advanced enough that they may
soon be a useful guide for policy making. McCallum's chapter is a big step in
the right direction. Analyses of this type should become the standard, by
which monetary policy is conducted and assessed in the future. And the
discussion should become more informed, moving away from the more
primitive IS±LM language to the more elegant language of applied general
equilibrium analysis.
Indeed, this is already happening in the United States. To an outside
observer, it is remarkable how far and how fast the various branches of the
Federal Reserve System in the United States have moved in building up their
expertise for analysing monetary and economic phenomena. By providing
suf®cient freedom and resources to actively participate in frontier research on
economics and monetary policy, and by encouraging publication in leading
journals and presentations of ®ndings at learned meetings, the research
departments at the board of the Federal Reserve System as well as at many
regional Federal Reserve Banks have provided an environment so attractive
that it has enabled them to recruit the best talent available. They now rival the
leading economics departments at US universities in quality. There is a
friendly, healthy competition between the various branches within the
Federal Reserve System, keeping the analysts at each place honest and working
hard to provide the best analysis possible, with the largest research department
at the centre of the system. The relevance for monetary policy decision making
is obvious: the weight and the clout of the members of the Board of Governors
of the Federal Reserve System increases directly with the weight and the clout
Discussion 59
of the research department to which they have access, with Alan Greenspan's
weight particularly large. For the public, the result has been the best monetary
policy that the United states has ever enjoyed. Obviously, one cannot ascribe
the entire success and the in¯uence of individual board members to the
strength of the research departments. But one shouldn't run the risk of
underestimating it either.
It is hard to imagine that a similar logic will not hold true for the European
System of Central Banks as well. Indeed, it rather seems that the race to the top
has already begun. The economists at these central banks will use the best
available tools to study the important monetary policy questions at hand.
They will therefore be students of McCallum's work for a long time to come.
References
Blinder, A. S. (1989) Central Banking in Theory and Practice, Cambridge, MA, MIT Press.
Cukierman, A. and A. H. Meltzer (1986) `A Theory of Ambiguity, Credibility, and
In¯ation under Discretion and Asymmetric Information', Econometrica, 54, 1099±
1128.
Fuhrer, J. (1997) `An Optimization-based Econometric Framework for the Evaluation of
Monetary Policy ± Comment', NBER Macroeconomics Annual 1997, 346±55.
Ghironi, F. (1999) `Alternative Monetary Rules for a Small Open Economy: The Case of
Canada', University of Berkeley, Department of Economics, November, draft.
Kydland, F. E. and E. C. Prescott (1977) `Rules Rather than Discretion: The Inconsistency
of Optimal Plans', Journal of Political Economy, 85, 473±92.
McCallum, B. T. and E. Nelson (1998) `Nominal Income Targeting in an Open-economy
Optimizing Model', NBER Working Paper, 6675; also in Journal of Monetary Economics,
43.
2
1 Introduction
60
De®ning and Maintaining Price Stability 61
instead means a situation with low and stable in¯ation, `low in¯ation'
(including zero in¯ation). The former de®nition implies that the price level is
stationary (or at least trend-stationary). The latter de®nition implies base drift
in the price level, so that the price level will include a unit root and be non-
(trend-)stationary. Indeed, the price-level variance increases without bound
with the forecast horizon. Thus, to refer to low in¯ation as price stability is
indeed something of a misnomer.
Let me refer to a monetary policy regime as price-level targeting or in¯ation
targeting, depending upon whether the goal is a stable price level or a low and
stable in¯ation rate. We can represent (strict)3 price-level targeting with an
intertemporal loss function
X
1
Et Lt;
2:1
0
to be minimised, where (0 < < 1) is a discount factor and the period loss
function is the quadratic loss function
1
Lt
pt � pt 2
2:2
2
Here, pt denotes the (log) price level in period t and pt denotes the (log) price-
level target. The price-level target could be a constant or a (slowly) increasing
path,
than from the target price level pt�1 . Similarly, (2.6) illustrates that the
in¯ation target becomes endogenous and time-varying under price-level
targeting.
In the real world, there are currently an increasing number of monetary
policy regimes with explicit or implicit in¯ation targeting, but there are no
regimes with explicit or implicit price-level targeting. Whereas the Gold
Standard may be interpreted as implying implicit price-level targeting, so far
the only regime in history with explicit price-level targeting occurred in
Sweden during the 1930s (see Fisher, 1934, and Berg and Jonung, 1999; this
regime was quite successful in avoiding de¯ation).
Even if there are no current examples of price-level targeting regimes, price-
level targeting has been subject to an increasing interest in the monetary
policy literature. At the 1984 Jackson Hole Symposium, Hall (1984) argued for
price-level targeting. Several recent papers compare in¯ation targeting and
price-level targeting, some of which are collected in Bank of Canada (1994).
Some papers compare in¯ation and price-level targeting by simulating the
effect of postulated reaction functions. Other papers compare the properties of
postulated simple stochastic processes for in¯ation and the price level (see
Fischer, 1994). A frequent result, which has emerged as the conventional
wisdom, is that the choice between price-level targeting and in¯ation
targeting involves a tradeoff between low-frequency price-level variability on
the one hand and high-frequency in¯ation and output variability on the
other. Thus, price-level targeting has the advantage of reduced long-term
variability of the price level. This should be bene®cial for long-term nominal
contracts and intertemporal decisions, but it would come at the cost of
increased short-term variability of in¯ation and output. The intuition is
straightforward: in order to stabilise the price level under price-level targeting,
higher-than-average in¯ation must be succeeded by lower-than-average
in¯ation. This would seem to result in higher in¯ation variability than under
in¯ation targeting, since base drift is accepted in the latter case and higher-
than-average in¯ation need only be succeeded by average in¯ation. Via
nominal rigidities, the higher in¯ation variability would then seem to result in
higher output variability.5
However, this intuition may be misleadingly simple. In more realistic
models of in¯ation targeting and price-level targeting with more complicated
dynamics, the relative variability of in¯ation in the two regimes becomes an
open issue. As shown in Svensson (1997b, appendix), this is the case if there is
serial correlation in the deviation between the target variable and the target
level ± for instance, if the price level displays mean reversion towards the price-
level target under price-level targeting and in¯ation displays mean reversion
towards the in¯ation target under in¯ation targeting. Svensson (1999e) gives
an example where the absence of a commitment mechanism and at least
moderate persistence in the Phillips curve imply that in¯ation variability
becomes lower under price-level targeting than under in¯ation targeting,
De®ning and Maintaining Price Stability 65
stability translates into a more gradualist policy. Thus, if in¯ation moves away
from the in¯ation target, it is brought back to target more gradually.
Equivalently, in¯ation-targeting central banks lengthen their horizon and
aim at meeting the in¯ation target further in the future. As further discussed in
Svensson (1999c), concerns about output-gap stability, simple forms of model
uncertainty and interest rate smoothing all have similar effects under in¯ation
targeting ± namely, a more gradualist policy. The Sveriges Riksbank has
explicitly expressed very similar views.9 The Chancellor's remit to Bank of the
England (HM Treasury, 1997) mentions `undesired volatility of output'.10 The
Minutes from Bank of England's Monetary Policy Committee (Bank of
England, 1999) are also explicit about stabilising the output gap.11 Several
contributions and discussions by central bankers and academics in Lowe
(1997) express similar views. Ball (1999) and Svensson (1998b) give examples
of a gradualist approach of the Reserve Bank of New Zealand. Indeed, a quote
from the ECB (European Central Bank, 1999 p. 47) also gives some support for
an interpretation with > 0, as well as some weight on minimizing interest
rate variability:
explicit in¯ation target (a point target or a range) may be more important than
whether the target (the mid-point of the range) is 1.5, 2 or 2.5 per cent.
A symmetric in¯ation target implies that in¯ation below the target is
considered equally bad as in¯ation the same distance above the target (which
is the case if in¯ation targeting is represented by a symmetric loss function like
(2.7)). This would seem to be a precondition for in¯ation expectations being
focused on the in¯ation target. A point target with or without a tolerance
interval would, from this point of view, be better than just a range. A range
would, in turn, be better than an asymmetric formulation like `below 2 per
cent'. These aspects may be particularly important when persistent de¯ation is
a possibility, of which recent developments in Japan remind us. A symmetric
in¯ation target would seem to be the best defence against persistent de¯ation
and against the appearance of de¯ationary expectations.
Interestingly, a price-level target may have special advantages relative to an
in¯ation target in avoiding persistent de¯ation, since an unanticipated
de¯ation which makes the price level fall below the price-level target will, if
the price-level target is credible, result in increased in¯ation expectations that
will, in themselves, reduce the real interest rate and stabilise the economy.15
The basic problem of maintaining price stability is thus to set the monetary
policy instrument (or instruments) so as to minimise the intertemporal loss
function (2.1) with the period loss function (2.7), subject to current
information about the current and future state of the economy and the
transmission mechanism.
The transmission mechanism is taken to be represented by a linear model in
state±space form
Xt1 X ut1
A t Bit
2:8
xt1jt xt 0
where Xt
t ; yt ; yt ; ::; 10 (where 0 denotes transpose) is a column vector of nX
predetermined variables (also called state variables), xt is a column vector of nx
forward-looking variables (also called non-predetermined variables), it is a column
vector of ni central bank instruments (also called control variables), ut1 is a
column vector of nX exogenous iid shocks with zero means and a constant
covariance matrix uu and A and B are matrices of appropriate dimensions. The
predetermined variables include in¯ation, output, potential output and other
variables. I use the convention that the last element of the vector of
predetermined variables is unity. This is a convenient way of allowing non-zero
means of the variables; the last column of A is then a function of these means.
Although the framework is general enough for handling multiple monetary
policy instruments I will, realistically, assume that there is only one
De®ning and Maintaining Price Stability 69
Lt Yt � Y 0 W Yt � Y
function under commitment that minimises (2.1) in period 0 (see Backus and
Drif®ll, 1986; Currie and Levine, 1993; and Soderlind, 1998). This reaction
function will be a linear function of the predetermined variables and the
predetermined Lagrange multipliers (shadow prices) of the forward-looking
variables,
it f Xt 't 2:9
for t 0 and X0 given, where f and ' are row vectors with nX and nx elements
(called response coef®cients, or reaction coef®cients), respectively. Further-
more, the multipliers t ful®l
X
t
�1
it f Xt ' M22 M21 Xt�
2:11
1
it f Xt 2:12
Even when there are forward-looking variables, many papers consider the
optimal reaction function under commitment over the class of reaction
functions (2.12) of the current predetermined variables only (mostly without
notifying the reader that this is a restriction).
The optimal reaction function under commitment is normally a function of
all the predetermined variables (and the lagged predetermined variables) and
is, in this sense, a rather complex construction. Consider also the class of
simple reaction functions, the class of linear reaction functions restricted to
being `simple' in the sense of having few arguments (for instance, some of the
elements of vector f and all the elements of vector ' are restricted to zero). A
typical simple reaction function is the much-discussed Taylor rule, where the
instrument responds only to current or lagged in¯ation and the output gap.
Let the optimal simple reaction function under commitment be the reaction
function in a particular class of simple reaction functions that minimises (2.1)
in period 0, given X0 :
An optimal reaction function under commitment is likely to be too
complex, in the sense of involving speci®c responses to a large number of
predetermined variables, to be veri®able. Therefore it is dif®cult to conceive of
De®ning and Maintaining Price Stability 71
it f Xt 2:13
Then, instead of the period loss function (2.7), the central bank is committed
to the new period loss function
1
Lt
it � it 2
2:14
2
Clearly, a trivial ®rst-order condition for minimising (2.14) is given by16
it it 2:15
where r is the average real interest rate, g and gy are the long-run response
coef®cients and
0 < 1 is a smoothing parameter.
Furthermore, we realise that under this commitment to a simple policy rule,
the central bank need no longer be forward-looking. It need only be forward-
looking once and for all in period 0, when it decides to which simple reaction
function it will commit. After that, it need never be forward-looking; to set the
instrument according to the prescribed reaction function, or it simply needs to
minimise the period loss function each period with the prescribed interest rate
target.
Although most of the current and previous discussion of monetary policy
rules is in terms of commitment to alternative instrument rules (see, for
instance, McCallum, 1997, and the contributions in Bryant, Hooper and
Mann, 1993 and Taylor, 1999), I do not believe that a commitment to an
instrument rule is either a practical or desirable way of maintaining price
stability, for several reasons.
First, there are overwhelming practical dif®culties in deciding once and for
all which instrument rule to follow. The optimal reaction function will involve
72 Lars E. O. Svensson
information (stock market crash, Asian crisis, Brazil ¯oating) and correspond-
ing sound judgemental adjustment.
Although alternative instrument rules can serve as informative guidelines
(see, for instance, the contributions in Taylor, 1999 ± or, with regard to the
performance of a Taylor rule for the Eurosystem ± Gerlach and Schnabel, 1998,
Peersman and Smets, 1998 and Taylor, 1998) and decisions ex post may
sometimes be similar to those prescribed by the simple instrument rules, they
are not a substitute for a decision making procedure for the central bank.
Interest rate targeting for the Eurosystem was indeed rejected by the European
Monetary Institute (EMI), the predecessor of the European Central Bank (ECB),
in European Monetary Institute (1997, p. 1) (with, arguably, not the most
exhaustive argument):
Forecast targeting
As a background, recall that, with a quadratic loss function and a linear model,
with the assumption of a known model and only additive uncertainty,
certainty-equivalence applies. The problem of minimising the loss function
can be separated into a deterministic problem involving conditional forecasts,
the conditional means of current and future variables and a stochastic problem
involving deviations between realised outcomes and conditional forecasts.
The solution to the deterministic problem is the same as to the stochastic
problem (see Chow, 1975, for models with only predetermined variables and
Currie and Levine, 1993, for models with forward-looking variables as well).
Thus, the discussion can focus on the deterministic problem involving
conditional forecasts.
74 Lars E. O. Svensson
For any variable t , let tjt for 0 denote the expectation Et t given
information in a ®xed period t. The information in period t includes the
information available about the state of the economy as well as about the
model, (2.8).20 Let jt denote the future path tjt ; t1jt ; t2jt ; ::: Consider the set
I t of given paths ijt
itjt ; it1jt ; ::: of instrument settings, for which there exist
bounded paths jt and yjt � yjt of future in¯ation and output gaps,
respectively. For each ijt 2 I t , let jt
ijt denote the corresponding path for
variable and y (yjt is taken to be exogenous), and call it the corresponding
conditional forecast (conditional on information in period t, ijt and the model
(2.8)). Let
n o
YT jt
ijt ; yjt
ijt � yjt jijt 2 I t
Thus, the problem of the central bank is to choose the path ijt ; such that the
resulting jt and yjt minimise (2.1) with (2.17). The ®rst element of ijt , itjt , is
then the appropriate instrument setting for period t; it . If there is no new
relevant information in period t 1, the instrument setting in period t will be
the second element in ijt . If there is new relevant information, that
information is used for solving the problem again in period t 1.22
This procedure thus involves making conditional forecasts of in¯ation,
output and the output gap for alternative interest rate paths, using all relevant
information about the current and future state of the economy and the
transmission mechanism. It involves making consistent assumptions about
De®ning and Maintaining Price Stability 75
exogenous and endogenous variables (for instance, that exchange rates and
interest rates are consistent with appropriate parity conditions). As discussed
further below, it also allows judgemental adjustments and extra-model
information (p. 77), as well as partially observable states of the economy
(p. 79). As discussed on p. 82, forecast targeting can even be adapted to take
non-linearities and model uncertainty into account, which both result in non-
additive uncertainty.
The procedure requires estimates of policy multipliers, the effects on the
conditional forecasts of changes in the instrument. The policy multipliers are
easily calculated in a simpli®ed model with only predetermined variables
X
�1
Xtjt AXt�1jt Bit�1jt A Xtjt A�1�s Bitsjt
2:20
s0
dXtjt
A�1�s B
2:21
ditsjt
Optimality criterion
What is the criterion for having found an optimal interest rate path and
corresponding conditional forecasts of in¯ation and the output gap? One
criterion can be formulated as follows. Suppose the central bank staff have
constructed a potential optimal combination of an interest-rate path and such
conditional forecasts. Consider a change dijt
ditjt ; dit1jt ; ::: in the interest-
rate path ijt . This will result in changes dXjt
0; dXt1jt ; dXt2jt ; ::: in the
predetermined variables, given by
X
�1
dXtjt
dXtjt ditsjt
s0
dits
Let djt and dyjt denote the corresponding changes in the in¯ation and output
forecasts (the output-gap forecast yjt is taken to be exogenous). A necessary
condition for optimality is then that the corresponding change in the
intertemporal loss function is non-negative ± that is
X
1 X
1
d Ltjt
tjt � dtjt
0 0
ytjt � ytjt dytjt 0
2:22
76 Lars E. O. Svensson
Given jt , yjt , yjt d;jt and dyjt ; as well as , and ; this expression is easily
checked, and a relatively easy way of making pairwise comparisons of
alternative interest rate paths and conditional forecasts. For instance, a delay
in an interest rate increase can be compared with an immediate increase, a
small increase now can be compared with a larger increase two quarters later,
etc.
Within the simple model in Svensson (1997a, 1999c), I have shown that the
®rst-order conditions for an optimum can be expressed as particularly simple
targeting rules (in the form of equations for the conditional forecasts of the
target variables). For instance, the in¯ation gap t1jt � and the output
gap ytjt � ytjt should be of the opposite signs and related according to
y c
t1jt � �
ytjt � ytjt ; 1
1 � c
t1jt � c tjt � ; 1
In practice, the decision making body may get far by just visually examining
alternative in¯ation and output-gap forecasts and then choosing the one that
is the best compromise between hitting the in¯ation target at an appropriate
horizon and avoiding output-gap stability. If done in a consistent way, this
will be equivalent to minimising the intertemporal loss function. Compared
to many other intertemporal decision problems that households, ®rms and
investors solve one way or another (usually without the assistance of a
substantial staff of economics PhDs), this particular decision problem does not
seem to be overly complicated or dif®cult.
Instrument assumptions
In the above discussion, the problem is to ®nd the appropriate interest rate
path ijt , which requires constructing conditional forecasts for exogenous
interest rate paths.
Forecasts for unchanged interest rates, where the interest rate path ful®ls
itjt it�1; 0
can be used as indicating `risks to price stability'. They are used by Bank of
England and Sveriges Riksbank to motivate changes in the interest rate and
their direction (see p. 81f.)
Conditional forecasts can also be constructed for given reaction functions,
in which case the interest rate path is endogenous and ful®ls
De®ning and Maintaining Price Stability 77
itjt f Xtjt
for a given f .23 Some central banks, notably the Bank of Canada and the
Reserve Bank of New Zealand, construct forecasts conditional on reaction
functions involving what Rudebusch and Svensson (1999a) call `responding to
model-consistent forecasts', and what Batini and Haldane (1999) call `forecast-
based' reaction functions. This implies an interest rate path ful®lling
where T > 0 is the forecast horizon (typically some six±eight quarters), and
sometimes f 0:
Among reaction functions involving responses to forecasts, it would seem
more natural that the forecast responded to is one for unchanged interest rates.
Indeed, under strict in¯ation targeting, the optimal instrument adjustment is
proportional to the deviation of the conditional forecast for unchanged interest
rate from the in¯ation target, as demonstrated in Svensson (1999d).24
Judgemental adjustments
A major advantage of forecast targeting relative to commitment to an
instrument rule is that it allows a systematic and disciplined way of
incorporating judgemental adjustment and extra-model information, by
`®ltering information through the forecasts'.
Given that every model of the transmission mechanism is an abstraction
and a simpli®cation, and given that there is considerable uncertainty about
the details of the transmission mechanism, monetary policy will never, it
seems, be able to rely on models and the information entering models alone.
There will always be an important role for additional extra-model information
and judgemental adjustments of the instrument. Never using such informa-
tion and judgement would neither be ef®cient nor incentive-compatible for
the decision making body of the central bank. At the same time, such
information and informal adjustment opens up monetary policy to arbitrari-
ness and potential abuse. Forecast targeting allows some system and discipline
in the use of extra-model information and judgemental adjustments.
Judgemental adjustments can be made in several ways in the framework
outlined here. One is in the form of adjustments of the coef®cients of the
model. This means that the coef®cients of matrices A and B become time-
varying, Atjt and Btjt , which is easily incorporated when constructing the
forecasts (as long as the time-variation is deterministic; see p. 84 on non-
additive uncertainty).
Another kind of judgemental adjustments consists of simple additive shifts
in the forecasts. For the model with only predetermined variables, this means
that the model in period t can be represented as
That is, in period t, the central bank knows the current and past indicators, in
addition to the model and the stochastic properties of the shocks.
This formulation allows some variables to be directly observable, some to be
observed with measurement error and some to be completely unobservable.
The discussion in previous sections, the full information case, corresponds to
the special case Zt
X0t ; x0t ; i0t 0 .
For simplicity, assume that there are no forward-looking variables, so that
the model is (2.19), and let the indicators depend on the predetermined
variables according to
Zt DXt vt
2:23
This setup is examined in more detail in Chow (1975), Tinsley (1975) and
LeRoy and Waud (1977).30 The case of partial information with forward-
looking variables is examined in Pearlman, Currie and Levine (1986),
Pearlman (1992), Aoki (1998) and Svensson and Woodford (2000). The
present discussion follows Svensson and Woodford (2000) (although without
forward-looking variables).
Under these assumptions, certainty-equivalence continues to hold. In
period t, the central bank needs to form the estimate Xtjt of the predetermined
variables in order to construct its conditional forecasts and set its instrument.
The optimal estimate is given by a Kalman ®lter, with the updating equation
so the matrices K and I � KD give the weights on the indicators Zt and the
prior information Xtjt�1 . If D I and vv 0, Zt coincides with the
predetermined variables; then K I and all weight is on the indicators and
none on the prior information. Generally, a row in K gives the optimal
weights on the indicators in estimating the corresponding predetermined
variable. A column in K gives the weights given to the corresponding
indicator in estimating the different predetermined variables. Assume that a
particular indicator, Zjt , is equal to one of the predetermined variables, Xkt ,
plus a measurement error, vjt ;
De®ning and Maintaining Price Stability 81
If the variance of the measurement error approaches zero, all the elements in
row k of K approaches zero except element j, which approaches unity. Thus, all
the weight in estimating Xkt falls on Zjt . If the variance of the measurement
error goes to in®nity, Zjt becomes a useless indicator. The elements in column j
of K then all become zero. The indicator Zjt gets zero weight in estimating the
predetermined variables.
Assume, for simplicity, that forecast targeting has resulted in a reaction
function f in the past. That is, it� f Xt�jt� for 1 t. Then, by (2.9),
Xj�1 A Bf X�1j�1
This gives the weight on indicators Zt� as
1 � KD
A Bf K for 0.
These equations illustrate the gradual updating of the estimate of the
predetermined variables. We can summarise the effects of the indicators in
period t on the forecasts Xtjt in terms of `indictor multipliers,' dXtjt =dZt ,
given by
dXtjt dXtjt dXtjt
A K
dZt dXtjt dZt
Thus, the indicator multiplier is the product of the effect of the estimate of the
current state of the forecast, dXtjt =dXtjt , which by (2.20) is given by A , and
the effect of the indicators on the estimate of the current state, with by (2.24) is
given by the gain matrix K. It follows that an indicator will affect the
instrument setting via affecting the current state of the economy, then the
forecasts, and ®nally the instrument. Schematically,
Zt ! Xtjt ! Xtjt ! it
the in¯ation target, ± should be useful when discussing monetary policy that
aims to maintain price stability. What requirements should such an indicator
ful®l? It would seem that, ®rst, it should signal in which direction and to what
extent the in¯ation target will be missed if policy is not adjusted. Second, it
should signal in which direction and to what extent the instrument should be
adjusted. Finally, it should be intuitive and easy to understand, so that it can
be used to communicate with the public and explain why an instrument
change is warranted or not.
The Eurosystem has put forward a money growth indicator, namely, the
deviation between current M3 growth and a reference value as an indicator of
risks to price stability - indeed, the ®rst of the `two pillars' of its monetary
strategy. As discussed in Svensson (1999d), such a money growth indicator
seems quite unsuitable for this purpose, since it is largely just a noisy indicator
of the deviation of current in¯ation from the in¯ation target (which deviation
can be more easily observed directly).32 Instead, the obvious candidate is a
conditional in¯ation forecast, conditional upon unchanged monetary policy
in the form of an unchanged interest rate. That is, it is constructed for the
interest rate path that ful®ls itjt it�1 , 0. Constructing this in¯ation
forecast is straightforward in a model without predetermined variables, as is
apparent from (2.20). It is somewhat more complicated in a model with
forward-looking variables, as shown in Svensson (1998a Appendix A).33 This
indicator signals whether and in which direction the in¯ation target is likely to
be missed, if policy is not adjusted, and it thereby also signals in which
direction the instrument needs to be adjusted.
The Bank of England and the Sveriges Riksbank, in their quarterly In¯ation
Reports, use conditional in¯ation forecasts for unchanged interest rates as their
main vehicle for motivating why the instruments need to be adjusted or not.
Non-linearities
So far, the maintained assumption has been that the model is linear. Sources of
non-linearity that have been discussed in the literature include non-linear
Phillips curves (see Debelle and Laxton 1997; Gordon, 1997; Isard and Laxton,
1996), non-negativity of nominal interest rates and downward nominal rigidity
of prices and/or wages (see references on pp. 66±8). Suppose now that the model
is non-linear. Assume that the model remains known, that there are no
De®ning and Maintaining Price Stability 83
2
X
1
Et
t � dt Et
yt � yt
dyt
0
X
1 n o
tjt � dtjt
ytjt � ytjt
dytjt
0
X
1
Covt t; dt Covt yt; � yt;
dyt
2:27
0
dXt dXtjt
dXt ditsjt ditsjt A�1�s B ditsjt
dits ditsjt
84 Lars E. O. Svensson
where we use that dXt =dits dXtjt =ditsjt A�1�s B, so dXt is determi-
nistic and independent of Xt (since the policy multipliers are constant).
With the non-linear model 2.26, the same change is
!
Y @M
Xtsr; itsr; utsr1 @M
Xts; its; uts1
r�s�1
dXt ditsjt
r1
@X @i
so that dXt is stochastic and generally correlated with Xt (since the policy
multipliers dXt =ditsjt are now endogenous and not constant).35
With a non-linear model, the optimal reaction function is nonlinear,
it f Xt
and it will generally depend on the covariances uu . The covariance terms in
(2.26) also imply that the optimal policy may imply a bias, in the sense that it
is optimal to, on average, either overshoot or undershoot the in¯ation target
(also, the optimal average output gap may not be zero).
Doing forecast targeting under a non-linear model and selecting the
instrument such that (2.17) is minimised would still imply a non-linear
reaction function, since the conditional forecasts would be non-linear
functions of the instrument path. It would imply disregarding the optimal
bias, though. How costly forecast targeting would be relative to the optimal
policy would, of course, depend on the degree of non-linearity in the
transmission mechanism, which is an empirical question. My reading of the
literature is that there is considerable controversy about the extent and
the relevance of any nonlinearity (see for instance Gordon, 1997, and Isard
and Laxton, 1996).
Non-additive uncertainty
So far, only additive uncertainty has been considered, appearing as the
additive shock ut1 in (2.8). Uncertainty about parameters in the model ± that
is, uncertainty in the coef®cients of the matrices A and B in (2.8) ± results in
multiplicative uncertainty, an example of non-additive uncertainty. Multi-
plicative uncertainty has consequences similar to non-linearity, in that
certainty-equivalence no longer holds, even if the model remains linear and
the loss function is quadratic.
Assume that the model is linear, that there are no forward-looking variables
and that the predetermined variables are observable, but assume now that the
model has time-varying stochastic parameters, with known stochastic
properties. Then the model can be written
! !
Y
�1 X
�1 Y
�s�1 X
Y
�s
Xt At1s Xt At1sr Bts1 its Atsr uts
s0 s0 r1 s1 r1
Intermediate targeting
When would intermediate targeting be optimal? Assume, for simplicity, strict
in¯ation targeting, (2.4). Consider, for simplicity, the model without forward-
looking variables, (2.19). Decompose the vector of predetermined variables
according to Xt
t ; X02t ; X03t 0 , where the ®rst element is in¯ation and the rest
of the variables are decomposed into two vectors, X2t and X3t . Suppose that
De®ning and Maintaining Price Stability 87
the two vectors X2t and X3t can be chosen such that the model (2.19) of the
transmission mechanism ful®ls
2 3 2 32 3 2 3
t1 0 A12 0 0
4 X2;t1 5 4 A21 A21 A23 54 X2t 5 4 B2 5it ut1
X3;t1 A31 A32 A33 X3t B3
that is, where the A and B matrices are such that A11 0, A13 0 and B1 0.
Then in¯ation ful®lls
and is exclusively determined by variables X2t ; and variables X2t are the only
predictors of in¯ation (aside from the zero-mean exogenous shock u1;t1 ).
Under these assumptions, the instrument it affects in¯ation exclusively by ®rst
affecting X2;t1 and then by X2;t1 affecting t2 . Schematically, we have
it ! X2;t1 ! t2
Because of this property of X2t ; its elements can be called intermediate variables.
Let X2 ful®ll
Substituting (2.29) and (2.30) into (2.4) for t 1 and taking the expectation in
period t result in
2
where u1 is the variance of u1t and the weight matrix W ful®ls W 12 A012 A12 .
Clearly, using the period loss function
is equivalent to using (2.4). Now we can call X2t intermediate target variables, X2
intermediate target levels, and minimising (2.1) with L~ t instead of (2.7) we can
call intermediate targeting. We thus have a situation where intermediate
targeting is as good as strict in¯ation targeting.
In particular, assume that in¯ation is exclusively determined by money
growth according to
where mt mt � mt�1 and mt is the log of a monetary aggregate. Then, the
instrument exclusively affects in¯ation via ®rst affecting money-growth ± that
is,
it ! mt1 ! t2
88 Lars E. O. Svensson
Thus, let X2t mt , A12 1, X2 and W 12 ; and strict in¯ation
targeting can be replaced by strict money growth targeting with the period
loss function
1
L~ t
mt � 2
2
Both kinds of targeting will be equivalent.
In the example above, the transmission mechanism is recursive in a special
way, such that the target variables (in the above case only in¯ation) are
determined only by a set of intermediate variables (the only exception being
zero-mean exogenous shocks). Clearly, this is an extremely special case. In the
real world, and in reasonable models, the transmission mechanism is too
complex for intermediate variables in this sense to exist ± that is, the
transmission mechanism is not recursive in the above sense.38
Therefore, intermediate targeting in general, and monetary targeting in
particular, is not a good monetary policy strategy. However, there is one
exception to the general non-existence of intermediate variables. As discussed
in Svensson (1997a, 1999b), one set of intermediate variables always exists,
namely conditional forecasts. For any vector Yt of target variables, we can write
. `a prominent role for money with a reference value for the growth of a
monetary aggregate' and
. `a major role for a broadly-based assessment of the outlook for future price
developments'.39
With regard to the role of money, the Eurosystem has emphasised that the
reference value should not be interpreted as an intermediate target for money
growth. Indeed, it has rejected monetary targeting, on the grounds that the
relationship between money and prices may not be suf®ciently stable, and that
it is not clear that the monetary aggregates with the most stable relationship is
suf®ciently controllable in the short run. As Issing (1998) summarises:
As I argue in some detail in Svensson (1999d), there is little ground for such a
prominent role for money. It is easily shown in the simple and conventional
model used there, that such a money growth indicator will be a relatively useless
indicator of risks to price stability ± and, indeed, mostly a noisy indicator of the
deviation of current in¯ation from the in¯ation target. As argued on p. 79±82, the
weight on money as an indicator should be strictly determined by its predictive
power in forecasting in¯ation. As argued on p. 81±82 and demonstrated in some
detail in Svensson (1999d), the best indicator of `risks to price stability' is an
in¯ation forecast conditional on an unchanged interest rate.40
I believe it worthwhile to look more closely at the Eurosystem's arguments
in favour of a prominent role to money, stated very clearly in Issing (1998).
Three arguments for giving a prominent role for money are provided: (1)
`In¯ation is fundamentally monetary in origin over the longer term', (2) `It
creates a ®rm `nominal' anchor for monetary policy and therefore helps to
stabilise private in¯ation expectations at longer horizons', and (3) `[It]
emphasises the responsibility of the ESCB for the monetary impulses to
in¯ation, which a central bank can control more readily than in¯ation itself'.
With regard to argument (1), it is based on the empirical high long-run
correlation between money and prices. This correlation, however, holds in any
model where demand for money is demand for real money, for instance in the
simple model used in Svensson (1999d) to demonstrate the inferiority of the
Eurosystem's money growth indicator. The correlation is actually a relation
between two endogenous variables, and says nothing about causality. The
direction of causality is determined by the monetary policy pursued. Under
strict monetary targeting, when the central bank aims at maintaining a given
money growth rate regardless of what happens to prices, money growth
becomes exogenous in the relation and causes in¯ation, which is endogenous.
Under in¯ation targeting, when the central bank aims at maintaining a given
in¯ation rate regardless of what happens to money, in¯ation becomes
exogenous in the relation and causes money growth, which is endogenous.
Hence, argument (1) is neutral to the monetary strategy.
With regard to argument (2), it seems that the de®nition of price stability
provides the best nominal anchor and is the best stabiliser of in¯ation
De®ning and Maintaining Price Stability 91
Certainly, forecasting and forecast targeting will not be easy, and forecast
uncertainty is likely to be relatively large. Nevertheless, forecasting is simply
necessary, given `the need for monetary policy to have a forward-looking,
medium-term orientation' that Issing and the Eurosystem emphasise.
Furthermore, forecast targeting implies using existing information in the
most ef®cient and ¯exible way. It incorporates both model and extra-model
information, allows judgemental adjustments, takes additive uncertainty for
granted and even allows imperfect understanding of the transmission
mechanism and model uncertainty, as I have tried to explain in this chapter.
Of course, the less the uncertainty and the better the understanding of the
92 Lars E. O. Svensson
5 Conclusions
This chapter argues that forecast targeting is the best way of maintaining price
stability, on the grounds that with lags and uncertainty in the transmission
mechanism, forecast targeting is the most ef®cient and ¯exible way of using
available information. By generalising forecast targeting from mean forecast
targeting to distribution forecast targeting, it should also be the best way of
handling model uncertainty. Indeed, I believe the current best practice in
central banks' maintaining price stability must be understood as forecast
targeting.
The chapter has, so far, discussed only the framework for policy decisions
and not at all the central bank's communication, degree of transparency and
degree of accountability. Under forecast targeting, the conditional forecasts for
in¯ation and the output gap are the crucial inputs in the policy decision.
Therefore, policy decisions are best explained and motivated, and policy is best
understood and anticipated by the public, with reference to these conditional
forecasts. This has the bene®cial effect that any criticism of the policy must be
more speci®c: for instance, is it the target or the central bank's forecast that is
wrong? Furthermore, making these forecasts public provides the best
opportunity for outside observers to monitor and evaluate the central bank's
policy, and making sure that its decisions are consistent with its objective.
Then policy can be evaluated almost in real time, without waiting some two
years to see the outcome of an in¯ation rate that is, by then, contaminated by
a number of intervening shocks. Finally, making the forecasts public provides
the strongest incentives for the central bank to improve its competence and do
the best possible job.
These are strong arguments in favour of making these forecasts public, a
practice already followed by the Reserve Bank of New Zealand, the Bank of
England, and the Sveriges Riksbank. Against this background, the Eurosys-
tem's refusal to publish its forecasts, citing far from convincing arguments,43 is
very dif®cult to understand, except perhaps as an expression of an initial lack
of con®dence and experience and a desire to further improve its competence
before going public (but if so, why not announce that the forecasts will
eventually be made public?). I see no reason why the Eurosystem should not
aim for the current best standard of transparency, as demonstrated by the
central banks already mentioned.
De®ning and Maintaining Price Stability 93
Notes
1. The paper which this chapter is based upon was presented at the Bundesbank
conference on `The Monetary Transmission Process: Recent Developments and
Lessons for Europe', 26±27 March, 1999. I thank the discussants Mervyn King and
Jose Vi~nals, and Claes Berg, Donald Brash, John Faust, Torsten Persson, Anders
Vredin and participants in seminars at IIES and Sveriges Riksbank for comments.
Special thanks are due to Jon Faust and Dale Henderson. Over the years, I have
bene®ted a great deal from many discussions with them on monetary policy and
from their very constructive (sometimes relentless) criticism of previous work of
mine. Dale has also directed me to an early, very relevant, literature. I also thank
Christina Lo È nnblad for editorial and secretarial assistance and Marcus Salomonsson
for research assistance. Needless to say, the views expressed and any errors are my
own responsibility.
2. This literature includes Bernanke et al (1998), Bernanke and Mihov (1997), Clarida,
Gali and Gertler (1998a), Clarida and Gertler (1997), Laubach and Posen (1997),
Neumann (1997), von Hagen (1995) (note a crucial typo: the coef®cient for money
supply in Table 1 should be 0.07 instead of 0.7).
3. `Strict' and `¯exible' targeting is de®ned below.
4. If arguments in favour of a small positive in¯ation rate is accepted, an upward-
sloping price-level target path may be preferable.
5. An interesting issue is to what extent the degree of nominal rigidity depends on
whether there is in¯ation or price-level targeting.
6. This result requires at least moderate output persistence with a Lucas-type Phillips
curve, and does not hold for a Lucas-type Phillips curve without persistence. Kiley
(1998) shows that the result does not hold for a Calvo-type Phillips curve without
persistence.
7. See, for instance, McCallum and Nelson (1999) and Williams (1997).
8. As in¯ation-targeting central banks, like other central banks, also seem to smooth
instruments, the loss function (2.7.) may also includes the term
it � it�1 2 with
> 0.
9. See box on p. 26 in Sveriges Riksbank (1997) as well as Heikensten and Vredin
(1998).
10. `actual in¯ation will on occasions depart from its target as a result of shocks and
disturbances. Attempts to keep in¯ation at the in¯ation target in these
circumstances may cause undesirable volatility in output'.
11. See Bank of England (1998), para. 40: `[I]n any given circumstances, a variety of
different interest rate paths could in principle achieve the in¯ation target. What
factors were relevant to the preferred pro®le of rates?... There was a broad consensus
that the Committee should in principle be concerned about deviations of the level
of output from capacity'.
12. The Reserve Bank's target was previously de®ned in terms of a somewhat complex
underlying in¯ation rate. In the Policy Target Agreement of December 1997, there
was a change to the more transparent CPIX.
13. On the other hand, the argument that in¯ation increases capital market
distortions, examined in Feldstein (1997, 1999), would, under the assumption of
unchanged nominal taxation of capital, motivate a zero or even a negative in¯ation
target.
14. For reasons explained in Gordon (1996), I believe that Akerlof, Dickens and Perry
(1996) reach too pessimistic a conclusion. On the other hand, their data is from the
94 Lars E. O. Svensson
United States and Canada, and downward nominal wage rigidity may be more
relevant in Europe. The conclusions of Orphanides and Wieland (1998) are
sensitive to assumptions about the size of shocks and the average real interest rate;
the latter is taken to be 1 per cent for the United States. If the average real rate is
higher in Europe, and the shocks not much larger than in the United States, non-
negative interest rates may be of less consequence in Europe. Wolman (1998a,
1998b) provides a rigorous examination of the consequences of non-negative
interest rates in a more explicit model, and ®nds relatively small effects.
15. Also, Wolman (1998b) ®nds that a reaction function responding to price-level
deviations from a price-level target (rather than in¯ation deviation from an
in¯ation target) has good properties for low in¯ation rates.
16. Note that this simple ®rst-order condition only arises if the variables in Xt are
predetermined.
17. There is an additional philosophical objection to once-and-for-all commitment:
how come the once-and-for-all commitment can be done in period 0? Why was it
not already done before, so nothing remains to be committed to in period 0? Why
is there something special about period 0?
18. I found this appropriate quote in Budd (1998).
19. See Budd (1998) for an interesting and detailed discussion of the advantages of
explicitly considering forecasts rather than formulating reaction functions from
observed variables to the instrument.
20. It is important that these expectations are conditional on the central bank's model,
and hence are `structural', rather than being private sector expectations, in order to
avoid the problems of non-existence or indeterminacy of equilibria, arising from
responding mechanically to private sector expectations, as has been emphasised in
Woodford (1994) and further discussed in Bernanke and Woodford (1997).
21. The conditional forecasts for arbitrary interest-rate path derived in Svensson
(1998a, Appendix A] assume that the interest rate paths are `credible' that is,
anticipated and allowed to in¯uence the forward-looking variables. Leeper and Zha
(1999) discuss an alternative way of constructing forecasts for arbitrary interest rate
paths, by assuming that these interest rate paths result from unanticipated
deviations from a normal reaction function.
22. The consequences of imposing the restriction of time-consistency of it remain to be
examined. That is, that the elements itjt in it shall be consistent with the decision
in period t conditional on Xtjt (see n. 6, p. 93).
23. Note that one way of taking the discretionary nature of decision making into
account is to set it in period t under the restriction that the reaction function that
will apply in period t for 1 will be itjt ftjt Xtjt , where ftjt is the
reaction function that is likely to result from the decision in period t .
24. Note that an equation like it gXtTjt , where XtTjt is a model-consistent forecast
(including this equation), especially in a model with forward-looking models, is a
rather complex equilibrium condition. For reasons detailed in Svensson (1999b,
section 2.3.1), I am rather sceptical about these equilibrium conditions as reaction
functions representing in¯ation targeting.
25. See Tinsley (1975) and Reifschneider, Stockton and Wilcox (1997) for further
discussion of judgemental adjustments.
26. Mervyn King has emphasised that it is important that the decision making body of
the central bank agrees with the forecast. This requires iterations between the staff
and the decision making body, with the decision making body having the last say
on the forecast.
De®ning and Maintaining Price Stability 95
27. Thus, for the `strict' case with in¯ation as the only argument in the loss function,
the four loss functions are (1) Lt 12
t � 2 , (2) Lt jt � j, (3)
Lt k �
t � , where
x is the so-called R 1Dirac delta function with the
properties
x 0 for x 6 0;
0 1; and �1
xdx 1, and (4) Lt 0 for
jt � j a > 0, Lt k > 0 for jt � j > a.
28. Another problem with reporting the mode forecast is evident in the hypothetical
case when the forecast is bimodel with approximately equal probability density at
the two modes.
29. Two separate arguments are sometimes presented in favour of emphasising the
mode forecast. First, in presenting and discussing the forecast, it may often be natural
and intuitive to consider a most likely scenario together with one or two alternative
scenarios. The most likely scenario would then correspond to the mode forecast.
Second, before that stage, in constructing the forecast, it may be practical to start
with a most likely scenario and then add various uncertainties and complications
later on. Whereas the ®rst argument may be a legitimate argument in favor of the
mode, the second is not, since the presentation and the construction can be
independent.
Furthermore, the mode and the median have the property that they are not affected
by outlines, which may or may not be an advantage, depending on one's view.
30. See Orphanides (1998) and Smets (1998) for recent related work.
31. In a steady state, the Kalman gain is given by K PD0
DPD0 vv �1 , where the
covariance matrix P of the forecast errors Xt � Xtjt�1 is given by
P MP � PD0
DPD0 vv �1 DPM 0 uu , where M is the transition matrix in the
transition equation Xt1 MXt ut1 :
32. That such a money growth indicator is unsuitable on its own is fairly obvious, since
money is not the only, not even the major, predictor of in¯ation at the horizons
relevant for monetary policy (see Estrella and Mishkin, 1998). What is perhaps less
obvious is that the money growth indicator is unsuitable even for a completely
stable money-demand function without velocity shocks (see Svensson, 1999d, and
Rudebusch and Svensson, 1999b).
33. More speci®cally, with forward-looking variables, the interest rate is kept
unchanged for a few periods (four±six quarters, say), but then, there is a shift to
`normal' policy, or to some policy stabilizing in¯ation and determining the future
forward-looking variables.
34. A non-quadratic loss function would also imply that certainty-equivalence no
longer holds (see p. 78±79).
35. Note that we use the convention that tr s r 1 for t < s:
36. Recent work on and discussion of monetary policy under model uncertainty
includes Blinder (1998), Cecchetti (1997), Clarida, Gali and Gertler (1998b), Estrella
and Mishkin (1998), Levin, Wieland and Williams (1999), McCallum (1997),
Onatski and Stock (1998), Peersman and Smets (1998), Rudebusch (1998), Sack
(1988), Sargent (1998) Smets (1998), So È derstro
and the output gap are unlikely to be independent, distribution forecast targeting
requires the joint distribution to be conveyed. This may require some further
innovation in display, beyond the already beautiful fan chart.
38. As discussed in Wallis (1999), the Bank of England's fan charts present prediction
intervals that differ from normal con®dence intervals, central prediction
intervals. The Sveriges Riksbank, however, presents normal con®dence intervals,
(see Blix and Sellin, 1998). Both banks present the mode as their point forecast,
whereas it seems to me that it would be more natural and consistent with the
theory to present the mean (or, in distribution forecast targeting, at least the
median).
39. See Bryant (1980), Friedman (1975), Kalchbrenner and Tinsley (1975) and Kareken,
Muench and Wallace (1973) for this and other arguments against intermediate
targeting in general and monetary targeting in particular. See Rudebusch and
Svensson (1999b) for simulations of monetary targeting in the U.S. with lessons for
the Eurosystem. These simulations show that monetary targeting in the United
States would be quite inef®cient compared to ¯exible in¯ation targeting, in the
sense of bringing higher variability of both in¯ation and the output gap.
40. Since the ®rst version of this paper was written, an extensive discussion and
motivation of Eurosystem strategy has been presented by Angeloni, Gaspar and
Tristani (1999).
41. Rudebusch and Svensson (1999b) examine monetary targeting in an empirical
model of in¯ation, output and money for US data and draw some lessons for the
Eurosystem. They ®nd that monetary targeting would be very inef®cient compared
to in¯ation targeting, in the sense of increasing the variability of both in¯ation and
output. Counter to conventional wisdom, this is the case also if money demand
shocks are set to zero so the money demand is completely stable.
Gerlach and Svensson (1999) examine the indicator properties of monetary
aggregates for the euro area. Somewhat surprisingly, they ®nd considerable
empirical support for the so-called P model of Hallman, Porter and Small
(1991), adapted to Germany by To È dter and Reimers (1994). This implies that
monetary aggregates, in the form of the `real money gap', the gap between current
real balances and long-run equilibrium real balances, has considerable predictive
power for future in¯ation. They ®nd little or no empirical support for the
Eurosystem's money growth indicator, though.
Indeed, the theoretical analysis in Svensson (2000a) shows that the P model,
although emphasising the role of the real money gap in forecasting and controlling
in¯ation, does not provide any support for a Bundesbank-style money growth
target or a Eurosystem-style money-growth indicator.
42. See also Angeloni, Gaspar and Tristani (1999).
43. As I argue in Svensson (1999d), it may be sobering to recall that the introduction of
in¯ation targeting in the United Kingdom, Sweden and Finland occurred under
rather dramatic circumstances. The countries went through dramatic boom±bust
experiences, very serious banking and ®nancial sector crises, and a dramatic sudden
shift from a ®xed exchange rate to a new monetary policy regime with a ¯oating
exchange rate. Furthermore, this occurred in a situation with very low credibility
for monetary policy, with high and unstable in¯ation expectations, much above
the announced in¯ation targets. At least for Sweden (where I am naturally more
informed) the central bank's commitment to the ®xed exchange rate was so strong,
that there was no contingency planning. When the krona was ¯oated in November
1992, the new in¯ation-targeting regime, which was announced in January 1993,
De®ning and Maintaining Price Stability 97
had to be conceived from scratch (although, of course, with the bene®t of the
experiences mainly from New Zealand and Canada). It is not easy to rank
dif®culties and uncertainty about the transmission mechanism, but it seems to me
that the dif®culties facing the Eurosystem are still not of the same magnitude as the
dif®culties that the central banks of the United Kingdom, Sweden and Finland were
facing. Since those central banks have, nevertheless, managed quite well, the odds
for the Eurosystem may be quite good, provided it adopts a similar framework for
policy decisions.
44. See, for instance, Duisenberg (1998): `publishing an in¯ation forecast would
obscure rather than clarify what the Governing Council is actually doing. The
public would be presented with a single number intended to summarise a thorough
and comprehensive analysis of a wide range of indicator variables. However, such a
summary would inevitably be simplistic. Moreover, because publishing a single
in¯ation forecast would be likely to suggest that monetary policy reacts
mechanistically to this forecast, publication might mislead the public and therefore
run counter to the principle of clarity'.
References
Akerlof, G. A., W. T. Dickens and G. L. Perry (1996), `The Macroeconomics of Low
In¯ation', Brookings Papers on Economic Activity, 1, 1±76.
Angeloni, I., V. Gaspar and O. Tristani (1999) `The Monetary Policy Strategy of the ECB',
ECB Working Paper.
Aoki, K. (1998). `On the Optimal Monetary Policy Response to Noisy Indicators',
Princeton University, Working Paper.
Backus, D. and J. Drif®ll (1986) `The Consistency of Optimal Policy in Stochastic
Rational Expectations Models', CEPR Discussion Paper, 124.
Ball, L. (1997) `Ef®cient Rules for Monetary Policy', NBER Working Paper, 5952.
ÐÐÐÐ (1999) `Policy Rules for Open Economies', in Taylor (1999).
Bank of Canada (1994) Economic Behavior and Policy Choice under Price Stability, Ottawa,
Bank of Canada.
Bank of England (1998) Minutes of the Monetary Policy Committee meeting on 4±5 March
1998, Bank of England.
Batini, N. and A.G. Haldane (1999) `Forward-looking Rules for Monetary Policy', in
Taylor (1999).
Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience
1931±1937', Journal of Monetary Economics, 43, 3, 525±51.
Bernanke, B.S., T. Laubach, F.S. Mishkin and A.S. Posen (1998) In¯ation Targeting: Lessons
from the International Experience, Princeton University Press.
Bernanke, B.S., and I. Mihov (1997) `What Does the Bundesbank Target?', European
Economic Review, 41, 1025±54.
Bernanke, B.S., and F.S. Mishkin (1997) `In¯ation Targeting: A New Framework for
Monetary Policy?', Journal of Economic Perspectives; 11, 97±116.
Bernanke, B.S. and M. Woodford (1997) `In¯ation Forecasts and Monetary Policy',
Journal of Money, Credit, and Banking, 29, 653±84.
Blinder, A.S. (1998) Central Banking in Theory and Practice, Cambridge, MA, MIT Press.
Blix, M. and P. Sellin (1998) `Uncertainty Bands for In¯ation Forecasts', Sveriges Riksbank
Working Paper Series, 65.
Brainard, W. (1967) `Uncertainty and the Effectiveness of Policy', American Economic
Review, Papers and Proceeding, 57, 411±25.
98 Lars E. O. Svensson
Britton, E., P. Fisher and J. Whitley (1998) `The In¯ation Report Projections: Under-
standing the Fan Chart', Bank of England Quarterly Bulletin, 38, 30±7.
Bryant, R.C. (1980) Money and Monetary Policy in Interdependent Nations, Washington
D.C, Brookings Institution.
Bryant, R.C., P. Hooper and C.L. Mann (eds) (1993) Evaluating Policy Regimes: New
Research in Empirical Macroeconomics, Washington, DC, The Brookings Institution.
Budd, A. (1998) `Economic Policy, with and without Forecasts', Bank of England Quarterly
Bulletin, 86, 379±84.
Cecchetti, S.G. (1995) `In¯ation Indicators and In¯ation Policy', NBER Macroeconomics
Annual, 10, 189±219.
ÐÐÐÐ (1997) `Central Bank Policy Rules: Conceptual Issues and Practical Considera-
tions', Federal Reserve Bank of New York, Working Paper.
Chow, G.C. (1975) Analysis and Control of Dynamic Economic Systems, New York, John
Wiley.
Clarida, R., J. Gali and M. Gertler (1998a) `Monetary Policy Rules in Practice: Some
International Evidence', European Economic Review, 42, 1033±67.
ÐÐÐÐ (1998b) `The Science of Monetary Policy, A New Keynesian Approach', NBER
Working Paper 7147.
Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in
C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and Strategy, Chicago,
Chicago University Press.
Clark, P., D. Laxton and D. Rose (1996) `Asymmetry in the U.S. Output±In¯ation Nexus',
IMF Staff Papers, 43, 216±51.
Currie, D. and P. Levine (1992) `Should Rules Be Simple?', Economics and Planning, 25,
113±38, reprinted in Currie and Levine (1993).
ÐÐÐÐ (1993) Rules, Reputation and Macroeconomic Policy Coordination, Cambridge,
Cambridge University Press.
Debelle, G. and D. Laxton (1997) `Is the Phillips Curve Really a Curve? Some Evidence for
Canada, the United Kingdom, and the United States', IMF Staff Papers, 44, 249±82.
Duisenberg, W.F.(1998) `The ESCB's Stability-oriented Monetary Policy Strategy', speech
in Dublin, 10 November < http://www.ecb.int >.
Estrella, A. and F.S. Mishkin (1997) `Is there a Role for Monetary Aggregates in the
Conduct of Monetary Policy?', Journal of Monetary Economics, 40, 279±304.
ÐÐÐÐ (1998) `Rethinking the Role of NAIRU in Monetary Policy: Implications of
Model Formulation and Uncertainty', NBER Working Paper, 6518.
European Central Bank (1998a) `A Stability-oriented Monetary Policy Strategy for the
ESCB', Press Release, 13 October < http://www.ecb.int >.
ÐÐÐÐ (1998b) `The Quantitative Reference Value for Monetary Growth', Press
Release, 1 December < http://www.ecb.int >.
ÐÐÐÐ (1999) `The Stability-oriented Monetary Policy Strategy of the Eurosystem',
ECB Monthly Bulletin, January, < http://www.ecb.int >, 39±50.
European Monetary Institute (1997) The Single Monetary Policy in Stage Three: Elements of
the Monetary Policy Strategy in the ESCB, European Monetary Institute, Frankfurt.
Federal Reserve Bank of Kansas City (1996) Achieving Price Stability, Kansas City, Federal
Reserve Bank of Kansas City.
Feldstein, M. (1997) `The Costs and Bene®ts of Going from Low In¯ation to Price
Stability', in C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and
Strategy, Chicago, University of Chicago Press.
Feldstein, Martin (ed.) (1999) Costs and Bene®ts of Price Stability, Chicago, Chicago
University Press.
De®ning and Maintaining Price Stability 99
LeRoy, S.F. and R.N. Waud (1977) `Applications of the Kalman Filter in Short-run
Monetary Control', International Economic Review, 18, 195±207.
Lowe, P. (ed.) (1997) Monetary Policy and In¯ation Targeting, Sydney: Reserve Bank of
Australia.
McCallum, B.T. (1997) `Issues in the Design of Monetary Policy Rules', NBER Working
Paper, 6016.
McCallum, B.T. and E. Nelson (1999) `Nominal Income Targeting in an Open-economy
Optimizing Model', Journal of Monetary Economics, No. 43, 3.
Neumann, M. (1997) `Monetary Targeting in Germany', in K. Iwao (ed.), Towards More
Effective Monetary Policy, New York, St. Martin's Press.
Onatski, A. and J.H. Stock (1998) `Robust Monetary Policy under Model Uncertainty in a
Small Model of the US Economy', Harvard University, Working Paper.
Orphanides, A. (1998) `Monetary Policy Evaluation with Noisy Information', Federal
Reserve Board, Working Paper.
Orphanides, A. and V. Wieland (1998) `Price Stability and Monetary Policy Effectiveness
when Nominal Interest Rates Are Bounded at Zero', Federal Reserve Board, Working
Paper.
Pearlman, J.G. (1992) `Reputational and Nonreputational Policies under Partial
Information', Journal of Economic Dynamics and Control, 16, 339±57.
Pearlman, J.G., D. Currie and P. Levine (1986) `Rational Expectations Models with Partial
Information', Economic Modelling, 3, 90±105.
Peersman, G. and F. Smets (1998) `Uncertainty and the Taylor Rule in a Simple Model of the
Euro-Area Economy', Working Paper, presented at a conference on Monetary Policy and
Monetary Institutions held on March 5, 1999 under the joint sponsorship of the Federal
Reserve Bank of San Fransciso and the Stanford Institute for Economic Policy Research.
Reifschneider, D.L., D.J. Stockton and D.W. Wilcox (1997) `Econometric Models and the
Monetary Policy Process', Carnegie±Rochester Conference Series on Public Policy, 47, 1±37.
Rudebusch, G.D. (1998) `Is the Fed Too Timid? Monetary Policy in an Uncertain World',
Federal Reserve Bank of San Francisco, Working Paper.
Rudebusch, G.D., and L.E.O. Svensson (1999a) `Policy Rules for In¯ation Targeting', in
Taylor (1999).
ÐÐÐÐ (1999b) `Eurosystem Monetary Targeting: Lessons from U.S. Data', NBER
Working Paper, 7179.
Sack, B. (1988) `Does the Fed Act Gradually? A VAR Analysis', Federal Reserve Board,
Discussion Paper, 98±17.
Sargent, T.J. (1999) `Discussion of ``Policy Rules for Open Economies'' by Laurence Ball',
in Taylor (1999).
Smets, F. (1999) `Output Gap Uncertainty: Does it Matter for the Taylor Rule?' Bank for
International Settlements, Working Paper.
So
È derlind, P. (1998) `Solution and Estimation of RE Macromodels with Optimal Policy',
Stockholm School of Economics, Working Paper.
So È m, U. (1999a) `Monetary Policy with Uncertain Parameters', Stockholm School
È derstro
of Economics, Working Paper.
ÐÐÐÐ (1999b) `Should Central Banks Be More Aggressive?' Stockholm School of
Economics, Working Paper.
Stock, J.H. (1999) `Making Policies Robust to Model Uncertainty: Comment on ``Policy Rules
for In¯ation Targeting'' by Glenn Rudebusch and Lars E.O. Svensson', in Taylor (1999).
Svensson, Lars E.O. (1996) 'Commentary: How Should Monetary Policy Respond to
Shocks while Maintaining Long-run Price Stability?±Conceptual Issues', in Federal
Reserve Bank of Kansas City (1996).
De®ning and Maintaining Price Stability 101
Mervyn King
It has been said about a great newspaper that `all human life is there'. It could
equally be said about this chapter that `all Lars Svensson is there'. What
Chapter 2 does is to bring together many of Svensson's contributions to the
analysis of in¯ation targets, together with some new ideas, and ask the
question ± what can the European Central Bank (ECB) learn from this work? It
is a good question, and I hope that even if the ECB does not agree with
everything that Lars says, it will nevertheless bene®t from a serious discussion
of the points which he makes. I should like at the outset to acknowledge the
personal contribution which Lars Svensson has made to the theoretical
analysis of in¯ation targets. Those of us working in central banks with
in¯ation targets know only too well the debt we owe him for the intellectual
development which he has made. And, most important, this has been made in
a spirit of intellectual open-mindedness. This is important. No central bank
can claim a monopoly of wisdom, and in¯ation targets share much in
common with more traditional monetary policy frameworks such as monetary
targets. Both approaches can learn from the other. But it is clear that the 1990s
have been the decade of in¯ation targets, and Svensson has made a major
contribution to that achievement.
The central point in Svensson's chapter is the need for `a systematic
operational framework for policy decisions by central banks'. Svensson sees
forecast targeting as the most sensible way forward. Among the class of policy
frameworks in this category in¯ation targeting appears superior. Svensson
contrasts the systematic approach of forecast targeting with two alternatives.
These are, ®rst, simple instrument rules and, second, intermediate targets. I
agree with Svensson's ranking of these three alternative approaches to
monetary policy. But one must be a little cautious before drawing hard and
fast distinctions between them. No advocate of instrument rules would
suggest that policy be put on autopilot and that the rule be used in a
completely mechanical way to set policy. Equally, advocates of intermediate
monetary targets would wish, in certain circumstances, to deviate from the
policy implied by a mechanical application of those targets, as the Bundesbank
103
104 Mervyn King
did with money targets for many years. Of course, if the deviation from the
rule or the target is too great then its use either as a means of discipline or a
form of communication becomes low. But there is equally no simple
mechanical link between any particular summary statistic of the in¯ation
forecast and the choice of the policy instrument (usually the short-term
interest rate). There is always a judgement about what policy setting is
appropriate given the outlook for in¯ation.
The reason I prefer in¯ation targeting is that it makes clear that the
discretion used is in the mapping from the outlook for in¯ation to the choice
of instrument. That enables the central bank to divide its work into two parts.
The ®rst is to think through and then explain its assessments of the outlook for
in¯ation and output implied by the current state of the economy. The second
is to explain why, given that outlook, a particular policy choice was
appropriate. That seems to me to correspond to what central banks do in
practice. In contrast, simple instrument rules can be useful cross-checks on
choices made but are not very useful as communication strategies. And
intermediate targets likewise pose the problem of how one explains a
deviation of policy from that implied by the intermediate target. Surely that
would end up being an explanation in terms of the outlook for in¯ation.
Hence I like Svensson's phrase `distribution forecast targeting'.
But it is important to note that this is a framework for policy and not an
optimal rule for interest rates. And one factor which is striking in practice is
that the choice between the three approaches identi®ed by Svensson implies
nothing about views concerning the transmission mechanism of monetary
policy. Indeed, I could quite easily envisage my `optimal monetary policy
council', whether for the ECB or any other area, comprising individuals who
differed in their belief about the choice of policy framework. What matters
most is their ability to think of the transmission mechanism and to form
judgements about the current state of the economy.
Let me turn now to four speci®c points on Svensson's chapter. First,
Svensson advocates moving gradually from in¯ation targets to price level
targets. I have some sympathy with his arguments. There is still the important
question of how long is the horizon over which the price level is brought back
to its target path. It would be useful to carry out simulations to judge how far
an instrument rule, such as the Taylor rule, would imply a different interest
rate path if an error-correction term were added in order to bring the price level
back to a predetermined path within some reasonable time horizon. How
much greater would the volatility of interest rates be? But there is also a deeper
question of whether the costs of in¯ation are more to do with instability of the
expected price level over long horizons or of in¯ation itself. More work is
required on that question which, understandably, is not the focus of
Svensson's chapter.
Second, Svensson criticises the ECB for announcing a range for their
de®nition of price stability (close to an in¯ation target) rather than a point
Discussion 105
target. Again, I have some sympathy with this point. The symmetry of the
Bank of England's in¯ation target has been an immensely helpful aspect of our
remit because we have been able to show that it can be necessary to increase or
decrease interest rates according to circumstances. It is important to avoid
giving the impression that central banks are reluctant to cut interest rates and
quick to raise them. That would be the opposite of the politician's preference
for raising interest rates `when necessary' and lowering them `when possible'.
Symmetry should be an important part of the approach to setting interest
rates. The problem with a target range is that there is a lack of clarity about the
objective of monetary policy. Of course one might argue that the central bank
would do best by aiming for the centre of the range, and economic agents
would know that. The analogy is that, faced with an open goal and with some
uncertainty about your ability to kick a football, the best advice is to aim for
the point between the middle of the goalposts. Sadly, many footballers of my
acquaintance seem to be confused by an open goal, hesitate, and too often
miss. A point target has the virtue of concentrating the mind. It is also
particularly useful in the context of a policy making committee, to
demonstrate that the concept of `hawks' and `doves' on the same committee
has no meaning when the committee has a common point in¯ation target.
Third, Svensson suggests that the use of the reference value for money by
the ECB is misguided. The announcement of a money target might confuse
agents rather than `anchor' in¯ationary expectations. I have more sympathy
with the idea that money is special. There is still much that we do not fully
understand about the transmission mechanism, but it is dif®cult to talk about
in¯ation or monetary policy without according a special role for money. And
many of the models which play down the role of money and give a prominent
place to such concepts as the neutral real interest rate or the equilibrium
exchange rate presume more knowledge about the empirical values of these
concepts than is available in practice to central banks. So I think there are two
reasons for allowing a special role to money. The ®rst is that given uncertainty
about the appropriate model for the transmission mechanism, money may
have a robust property with respect to nominal outcomes. Second, there is a
real danger that focusing solely on traditional econometric models results in
trying to explain in¯ation solely in terms of real variables. So it is extremely
important that a central bank sees as one of its ®rst responsibilities the need to
explain movements in money and why it thinks that its policy stance is
consistent with the likely behaviour of money.
Fourth, Svensson considers the choice between targeting the mean, median
or modal future in¯ation rate. He points out that different loss functions
might imply the need to target different summary statistics of the distribution
of future in¯ation. More generally, one of the aspects of central bank
discretion is to decide how it should react to the entire distribution of in¯ation
prospects rather than commit itself to targeting a particular summary statistic
in all circumstances. Hence I much prefer Svensson's approach of considering
106 Mervyn King
the distribution of in¯ation as a whole and not just summary statistics such as
the mean, median or mode. That is also why in the `fan charts' for the in¯ation
outlook which are published in the Bank of England's In¯ation Report, we stress
that there can be no mechanical link between any particular line on that chart
and the policy decision. In some circumstances it may make sense to focus
very much on the mean outlook, and there is no doubt that over a long period
any deviation from that must cancel out if the target is to be met on average.
But there may be circumstances in which the mean is affected by a very small
probability of a signi®cant upside or downside risk. And it may not make sense
to adjust policy today to that eventuality. The fan chart should be constructed
in order to present information about the future outlook for in¯ation ± the
entire distribution ± in an informative way. The policy decision is separate.
In conclusion, I would stress that all central banks, whatever policy
framework they use, face essentially the same problem. The lags in monetary
policy mean that they must look forward to assess the impact of policy
decisions on the outlook for in¯ation and output. There are important
differences in the way in which the policy decision is presented, and the
choice of framework is not unimportant. As Svensson, I prefer the in¯ation
target approach. But I do believe that central banks can learn from each other,
and that just as the Bank of England learnt a great deal from the experience of
the Bundesbank in the period when it was developing its new approach and
acquiring the credibility which was necessary for the granting of independ-
ence in 1997, I hope that our experience will offer some lessons for the ECB.
Discussion
Jose Vin
Äals
107
108 Jose Vin
Äals
I must confess some vested interests in this topic. I ®rst discussed it in a paper
with Karl Brunner, more than thirty-®ve years ago (Brunner and Meltzer,
1963). I have returned to the topic many times, most recently in a published
symposium (Meltzer, 1995). I will refrain from reviewing these earlier studies,
although I will refer in passing to some of the main ideas. I will concentrate on
two topics. They do not exhaust the subject, but they raise issues that I believe
are central.
First, I raise some issues about a current class of models of monetary
transmission in which a short-term interest rate represents the transmission
process. The class of models is so widely accepted that the conclusions I
challenge have become part of the canon. A different class of models ± a more
useful one, I believe ± does more than give different answers. Some issues do
not arise; they are no longer relevant. And some issues remain relevant but
receive a different answer. The role of money is one such issue.
Second, I discuss some of the evidence I have gathered from my study ± A
History of the Federal Reserve ± the work that has been my main occupation for
the past four years. The two pieces are related, as I hope to show. The evidence
from history shows that the transmission process cannot be summarised by a
single interest rate. In the ®nal section, I present some econometric evidence
to supplement the historical data.
The core or central issue about the transmission of monetary policy is: How
does a monetary impulse affect relative prices and real demands as it moves
through the economy from its ®rst appearance to its ultimate effect on the
main determinants of economic welfare? The standard answer to this question
is, now as for some considerable time in the past, that monetary injections
change an interest rate in the money market. Because some prices, money
wages, or anticipations do not adjust instantly, there are effects on output and
employment that, though temporary, may be large and costly. Eventually,
112
The Transmission Process 113
for some, is a reality, (Krugman, 1988, Ito, 1998). The alternative denies that a
liquidity trap is possible except in the limit when all prices are zero.
A closely related proposition has received considerable attention as in¯ation
rates fell and remained low in the 1990s. Summers (1991) revived the
argument that a zero in¯ation target is socially costly because it sets a lower
bound for nominal interest rates. Monetary policy becomes weak or powerless;
it cannot lower the short-term nominal rate or prevent falling prices from
raising the real rate of interest. With money wages in¯exible downward,
unemployment rises. Akerlof, Dickens and Perry (1996) perform the remark-
able feat of ®nding evidence for this proposition using data for a period in
which in¯ation never remained close to zero. And Benhabib, Schmitt-Grohe
and Uribe (1998) argue that it is perilous to use a Taylor rule when in¯ation is
near zero.
A more sophisticated version of Summers' argument uses a stochastic model
with non-linearity in the transmission process when in¯ation is below 2 per
cent. Orphanides and Wieland (1998) ®nd that there is no evidence of an
operative lower bound in US postwar data. They claim that the lower bound
was in effect during the 1930s, so monetary policy was in¯exible for part of
that decade.
For this claim to be true, the short-term interest rate must be the principal or
only means by which monetary actions are transmitted from the central bank,
through the market, to the economy. As my old friend Karl Brunner often said:
we know this is false. Monetary actions are effective and powerful in the less
developed countries of Africa, Latin America, or Asia where there is no money
market. Relative prices respond to monetary impulses in countries without
central banks, and without money markets. There is more to the transmission
process than the models recognise.
Do these additional channels operate in countries with active money
markets? In writing the History of the Federal Reserve, I have found three
relevant examples from the years 1914 to 1951 that I have researched to date.
2 Historical evidence
1937±8
The National Bureau ranks the 1937±8 recession as the third most severe
recession in the years after the First World War. Real GNP fell 18 per cent and
industrial production 32 per cent in the thirteen months from May 1937 to
June 1938. Unemployment reached a peak of 20 per cent, not very different
from the 25 per cent peak in 1932.
The probable causes of the recession include both ®scal and monetary
actions. There is a very large reduction in the government de®cit in 1937 and a
very large reduction in growth of the monetary base. The main ®scal actions
are the end of the soldiers' bonus payment, the enactment of an excess pro®ts
tax to pay for part of the bonuses in ®scal 1937, and the start of social security
tax collections in ®scal 1936. The soldiers' bonus is the largest item,
$1.7 billion of current spending. It was paid in June 1936, in time for the
election later that year. The bonus was paid in bonds, but the bonds could be
sold for cash. By December 1936, $1.4 billion had been cashed. Balke and
Gordon's (1996) quarterly data show an 18 per cent average rate of increase in
real GNP for the last three quarters of 1936.
The most important monetary actions are the beginning of gold sterilisation
at the end of 1936 and the second and third increase in reserve requirement
ratios in March and May 1937. These increases completed the doubling of
reserve requirement ratios between August 1936 and May 1937. During the
entire period December 1936 to December 1938 that brackets the recession,
interest rates on Treasury bills remained between 0.03 per cent and 0.56 per
cent. Long-term nominal rates on Treasury bonds were modestly higher
during the recession than before or after, but the difference is small; the range
is 2.55 per cent ± 2.83 per cent.
Annualised monthly rates of price change are consistently negative from
October 1937 to February 1938 and intermittently negative for the rest of
1938. To smooth the data, I used moving twelve month averages of rates of
price change. Figure 3.1 compares the real interest rate to the annual growth
of the monetary base.
The common element in the two series is the twelve-month moving average
of the rate of price change. The divergence between the two series re¯ects some
release of sterilised gold into the monetary base in September 1937 and a small
per cent per cent
+7.5 +25
+6.0 +20
+4.5 +15
+1.5 +5
0 0
–1.5 –5
–4.5 –15
6.0
–6.0 –20
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1936 1937 1938
117
Figure 3.1 Year-over-year real base growth versus real long-term interest rate, January 1936±December 1938
118 Allan H. Meltzer
1948±9
The 1948±9 recession provides a second example refuting the liquidity trap
and the small or vanishing effect of monetary policy at low nominal interest
rates. The Federal Reserve pegged nominal long-term interest rates below the
2.5 per cent ceiling in effect from 1942 to 1951. Despite the pegging policy,
the monetary base fell through most of 1948. The principal reason is that the
Treasury used its budget surplus to retire debt held by the Reserve banks.
The monetary base fell as a consequence of the Treasury's actions. Although
the Federal Reserve complained about being an engine of in¯ation, prices fell
in half the months of 1948 and 1949.
The National Bureau dates the end of the expansion in November 1948 and
the recession trough in October 1949. The twelve-month moving average rate
of in¯ation fell from above 9 per cent in June and July 1948 to negative values
in May 1949. It remained negative for the rest of that year.
The Transmission Process 119
During most of the recession the Federal Reserve was more concerned about
a return of in¯ation than about the recession. The nominal rate on Treasury
bills remained between 1.02 per cent and 1.17 per cent throughout the
recession. Figure 3.2 compares annual growth of the real monetary base to the
real interest rate in the two years that include the recession. Data are computed
as in Figure 3.1. As before, the high positive correlation re¯ects the common
effect of the rate of price change on the two series. The high correlation and
parallel movement show that until late in 1949, when the recession was
almost over, the Federal Reserve took few actions to increase base growth.
Real base growth fell to ±11 per cent in September 1948, two months before
the cyclical peak. Thereafter base growth rose, but did not become positive
until April 1949, six months before the trough. The peak rate of base growth is
close to 6 per cent in August 1949, two months before the end of the recession.
At that time the real long-term interest rate was above 5 per cent.
Once again, the movement of real base growth is consistent with the
beginning and end of recession; the movement of real interest rates is not.
Once again, low nominal short-term interest rates do not appear to have
weakened the effects of monetary policy. And once again there appears to be
more to the transmission process than is contained in standard models.
1920±1
The third episode is the recession from January 1920 to July 1921.5 The
National Bureau ends the expansion in January 1920 and puts the last month
of recession in June 1921. The Federal Reserve undertook larger policy actions,
so nominal interest rates and nominal base growth re¯ect these actions.
In¯ationary policies in much of Europe and restrictive policies in the United
States brought an in¯ow of gold. The base and interest rate changes re¯ect
these in¯uences also.
Nevertheless, real base growth and real interest rates are positively correlated
during the recession. Both are negative at the start of the recession, turn
positive about a year later and reach a peak at the end of the recession. Judged
by base growth, monetary actions are countercyclical in the ®rst half of 1921;
judged by real interest rates, these actions are procyclical.
Figure 3.3 shows these data. The long-term nominal rate remains within a
narrow range but is higher at the trough of the recession than at the previous
peak. The dominant in¯uence on real rates and real base growth during the
recession is the decline in in¯ation followed by de¯ation.
As in the previous two episodes, interest rates give a misleading signal about
the thrust of policy. Real base growth gives a more correct signal. In this
recession, the de¯ation is severe; the peak annualised rate reached 17 per cent,
and it was above 10 per cent for ten consecutive months. The real long-term
interest rate, (i ± )/(1 ), is above 25 per cent at the end of the recession. The
economy recovered despite, not because of, the level of real interest rates.
120
per cent
per cent
+4.5 +6
+1.5 +2
0 0
–1.5 –2
–3.0 –4
–4.5
Twelve-month
–6
Twelve-month moving avarage real base growth
–6.0 Twelve month moving avarage real base growth –8
Twelve
–7.5 –10
–9.0 –12
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1948 1949
Figure 3.2 Twelve-month moving average real base growth versus real long-term interest rate, December 1947±December 1949
per cent
per cent
0
0
–15
Year-over-year real base growth –5
–30 –10
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1920 1921 1922
Figure 3.3 Year-over-year real base growth versus real long-term interest rate, January 1920±December 1922
121
122 Allan H. Meltzer
The three historical periods raise doubts about the central role assigned to
interest rates in the transmission process. They suggest an important role for
real balances. I return to these issues below.
An additional problem
Although the central bank's objective function is not directly part of the
transmission process, the three US recessions considered here raise doubts
about the choice of objective function used in current analyses. The usual
function has two arguments: (1) the difference between the in¯ation target
and the actual or expected in¯ation rate and (2) the output gap, the difference
between actual output and output consistent with the natural rate. Yet, we
have seen that faced with de¯ation and a deep recession (in three of the four
episodes) the Federal Reserve was slow to act. And it did not act effectively to
end de¯ation and recession. The objective function fails in these cases as a
positive statement of central bank objectives.6
My concerns about the standard objective function are not limited to its
empirical support, important as that is. I believe that using the output gap as
an objective of the central bank is problematic. This gap can arise for reasons
unrelated to monetary policy actions ± for example, an oil shock, reductions in
employment and output in the European Union resulting from provisions of
the welfare state, or other real events. My colleague Bennett McCallum
suggests that the problem can be overcome by rede®ning the natural rate to
take account of non-monetary effects (Chapter 1 in this volume). In principle,
this can be done; in practice it is dif®cult to do accurately.
The natural rate is not like the gravity constant. We neither measure it
precisely nor agree on its value. Opening the objective function to the pull and
tug of opinions about the size of the gap carries a risk to economic stability and
the apolitical position of an independent central bank. Recent discussion in
Europe and the United States shows that there can be differences of opinion
about measurement of both potential output and the natural rate of
unemployment.
A second problem with the now standard objective function is that it
neglects several issues of concern to central bankers. There is no role for a
lender of last resort. A run to currency, bank failures, a drain of foreign
per cent per cent
+10 +10
+5 +5
0
0
–5 Real M1 growth –5
–10 –10
Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1
Figure 3.4 Real M1 growth and real interest rate, August 1929±March 1993
123
124 Allan H. Meltzer
The three historical episodes discussed earlier suggest that an interest rate does
not fully represent the monetary transmission process. Changes in real money
balances appear at times to dominate changes in real interest rates as
indicators of the direction of change induced by monetary actions.
One interpretation of this evidence follows from the Haberler±Pigou±
Patinkin wealth effect. Falling prices raise private sector real money balances,
increasing real wealth. Some of the increase is spent on consumption. A
second interpretation views the change in real balances as a mixture of a pure
wealth effect and the changes in wealth induced by changes in the relative
prices of assets and output (Brunner and Meltzer 1963, 1968). On either
interpretation, we lose the simplicity of the two-equation, aggregate demand±
aggregate supply, class of models. If real balances affect aggregate demand, the
demand for money and perhaps other parts of the ®nancial structure become
relevant for the transmission process. The single interest rate is no longer
suf®cient to represent monetary policy and ®nancial markets.
The Transmission Process 125
Let me turn the issue around. The requirements of the standard model are
strong and are, as we have seen, easily falsi®ed empirically using historical
episodes. Analytically, the model is very demanding. Not only must the
wealth effect remain small and inconsequential for spending, but a single
interest rate must represent all relative price adjustment.
An appeal to rational expectations puts additional burdens on the model.
Market makers and participants must rationally anticipate how much output
prices will change to restore equilibrium at the natural rate. One data point, one
observation on the interest rate, is not suf®cient to determine whether the change
is permanent or temporary, a change in level or in growth rate, or an error that
will be reversed. Even if one believes that adjustment is rapid in markets for
actively traded assets, many assets are traded infrequently or not at all.
The transmission of monetary policy involves adjustment of prices of
existing hotels to newly produced hotels, existing plants and equipment to
newly constructed plants and equipment, new cars to used cars, new houses to
existing houses ± and, of course, money wages to prices.
I do not propose adding equations for each of these variables. That would
take us toward large-scale econometric models, a technology that has not
contributed greatly to our understanding of monetary policy. But there is also
considerable evidence that the expectations theory of the term structure or
interest arbitrage theories of the exchange rate are far from adequate. This
latter evidence suggests, again, that a single interest rate does not summarise
adequately responses on asset and output markets.
Recent analyses of the credit channel exploit differences in information to
develop an additional transmission channel for monetary actions (Bernanke
and Gertler, 1995). The credit channel operates parallel to the interest rate (or
monetary) channel, and the two interact. Empirical support for this channel
has not been persuasive; the hypothesised effect is probably swamped by the
endogenous response of bank loans to the monetary base and other
determinants of aggregate demand. The use of costly information to separate
parts of the transmission process is a step forward, however. It may ®nd a
better use distinguishing markets for bonds and real capital or distinguishing
different types of capital.
How well do real money balances capture the many channels of monetary
transmission? Koenig (1990) tested a two-stage model of changes in
consumption with changes in real money balances, real interest rates, income
and other variables as arguments of the consumption function. Consumption
includes only purchases of non-durables and services. The interest rate is
adjusted for in¯ation. All variables other than the interest rate and a dummy
variable are in logarithms. The dummy variable takes a value of unity in 1980
and after to represent ®nancial deregulation. All economic variables are used as
®rst differences.
Using data for the period from 1951:2 through 1986:1, Koenig found that
changes in real money balances had a signi®cant positive effect on changes in
126 Allan H. Meltzer
Table 3.1 Response of consumption to real balances and other variables, 1951:2±1986:1
Notes:
Standard errors in parentheses.
*Signi®cant at the 5 per cent level.
The Transmission Process 127
Notes:
Table 3.2 shows some results for the extended sample. The ®rst two columns
use changes in real M1 balances, as in Koenig's work. The last two columns
replace M1 with the St Louis monetary base, the variable I used in the
historical data after 1935. Other estimates using lagged values of money, real
income, stock prices and other variables give similar results. The only change
that made a substantial difference was omission of the dummy variable used to
shift the intercept after 1980.
These ®ndings suggest, again, that the transmission of monetary impulses
involves more than is found in the standard class of models. Changes in real
money balances, operating as a wealth effect or as a proxy for changes in both
relative prices and real wealth, have positive and signi®cant effects on the
change in consumption.
4 Conclusion
Viewed one way, the point of this chapter is an old one, one that is well-known
as the Haberler±Pigou±Patinkin real wealth effect. The conclusion in that case
would be that the wealth effect is more relevant in a dynamic context than
current, standard analysis recognises.
An alternative interpretation is that the dynamic real wealth effect includes
more than the direct effect of changes in real money balances. If anticipated
returns and anticipated in¯ation affect prices of assets with low transaction
costs more rapidly than the prices of new production, many changes in
relative prices of assets to output are part of the transmission process. As home
prices change relative to foreign prices, the real exchange rate changes. This
128 Allan H. Meltzer
Notes
* This chapter is a revised version of a dinner speech given by Allan Meltzer at the
conference.
1. My thanks to Randolph Stempski for his excellent assistance and to Bennett
McCallum for helpful discussion.
2. A similar but more complicated story is told about changes in money growth. Issues
about super neutrality are put aside.
3. I base my discussion on Svensson (1998). Other references could be used ± for
example, Taylor (1993, 1999).
4. A more complicated proof of this argument is in Brunner and Meltzer (1968).
5. For this period, the monetary base is high-powered money from Friedman and
Schwartz (1963). The price index is not seasonally adjusted.
6. Some may object that studies of Taylor's (1993) rule support use of the objective
function as currently relevant. Orphanides (1998) argues that this evidence is much
less compelling if we restrict the Federal Reserve to data available at decision time.
7. De®nition of variables is in Koenig (1990, pp. 422±4). We have used his de®nitions
wherever possible, but we are uncertain about his measurement of interest rates ±
end of quarter, quarterly average, etc. Interest rates are after tax rates, where tax rates
are marginal rates from Barro and Sahasakul (1983). We assumed constant marginal
tax rates after 1980.
References
Akerlof, G. A. Dickens W. T. and G. L. Perry (1996) `The Macroeconomics of Low
In¯ation', Brookings Paper on Economic Activity, 1, 1±76.
Balke, Nathan S. and R.J. Gordon (1996). `Appendix B: Historical Data' in R.J. Gordon
(ed.), The American Business Cycle: Continuity and Change. Chicago: University of
Chicago Press for the National Bureau of Economic Research, 789±842.
Barro, R. J. and C. Sahasakul (1983) `Measuring the Average Marginal Tax Rates from the
Individual Income Tax', Journal of Business, 56, 419±52.
Benhabib, J., S. Schmitt-Grohe and M. Uribe (1998) `The Perils of Taylor Rules', C. W.
Starr Center for Applied Economics, New York University, November, unpublished.
Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of
Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48.
Brunner, K. and A. H. Meltzer (1963) `The Place of Financial Intermediaries in the
Transmission of Monetary Policy', American Economic Review, 53, 372±82.
ÐÐÐÐ (1968) `Liquidity Traps for Money, Bank Credit and Interest Rates', Journal of
Political Economy, 76, 1±37.
Friedman M. and A. Schwartz (1963) A Monetary History of the United States, 1867±1960,
Princeton: Princeton University Press for the National Bureau of Economic Research.
Ito, T. (1998) `Japan and the Asian Financial Crisis: The Role of Financial Supervision in
Restoring Growth', paper presented at a conference on the `The Japanese Financial
System', New York, Columbia University, 1±3 October.
Kirchga ssner, G. and M. Savioz (1998) `Monetary Policy and Forecasts for Real GDP
Growth: An Empirical Investigation for the Federal Republic of Germany', University
of St Gallen, unpublished.
Koenig, E. F. (1990) `Real Money Balances and the Timing of Consumption', Quarterly
Journal of Economics, May, 399±425.
Krugman, P. (1998) `Japan's Trap', Krugman web page, MIT, May.
Meltzer, A. H. (1995) `Monetary, Credit, and (other) Transmission Processes', Journal of
Economic Perspectives, 9, 49±73.
130 Allan H. Meltzer
1 Introduction
The recent theoretical and empirical literature on monetary policy rules has
increasingly focused on short-term interest rates rather than monetary
aggregates for studying European monetary policy issues. There are several
reasons for this: ®rst, as in the United States, monetary aggregates in Europe
have displayed a less obvious link to real economic activity and in¯ation
during the 1980s and 1990s as opposed to the 1960s and 1970s. Second, many
central banks have de-emphasised the role of monetary aggregates and have
moved to operating procedures that focus more on interest rates (i.e. the Fed
funds target rate in the United States) or in¯ation rates (i.e. the in¯ation targets
in the United Kingdom, Canada, or New Zealand). Following the paper by
Taylor (1993) and more recent applications by Clarida and Gertler (1997),
Clarida, Gali and Gertler (1997, 1998), Gerlach and Smets (1998), Kuttner and
Posen (1998) and Rudebusch and Svensson (1998) there is now a growing
literature on so-called `interest rate smoothing' rules for Europe.1 These papers
use a simple policy reaction function in which interest rate adjustment
towards equilibrium depends on the deviations of in¯ation and output from
their respective target values. It is shown that such policy reaction functions ®t
the data quite well.
The present chapter takes a critical look at this new literature and discusses
two major problems. The ®rst of these problems is related to the interpretation
of such policy rules. It is argued that it is impossible to disentangle the causes
and consequences of policy actions in any policy reaction function, be it a
monetary policy rule or an interest rate policy rule, without imposing serious
prior restrictions. Unfortunately, such restrictions are typically dif®cult to
justify and to test. Think of the Clarida, Gali and Gertler (1998) paper, which
uses the level of short-term interest rates as the policy instrument and the
twelve-month ahead expected ( actual) in¯ation rate as well as the output
gap as the exogenous variables. If a restrictive monetary policy of increasing
interest rates in response to expected in¯ation is effective in easing
131
132 Axel A. Weber
France
16
14
a France
12
10
Belgium
8
6
4
Germany
Netherlands
2
0
1971 75 79 83 87 91 95 99
b Belgium
10
8
6
Netherlands
4
Austria
Belgium
Germany
2
1971 75 79 83 87 91 95 99
Italy
20
c
France
16
Netherlands
12
Italy
France
8
Germany
Germany
4
1971 75 79 83 87 91 95 99
Inflationtrates
(d) Inflation rates
26
22
Italy
18
14
France
10
Netherlands
6
Italy
2
Germany
–2
1971 75 79 83 87 91 95 99
for France, Italy and the remaining ERM countries have to be speci®ed differently
and this is why Clarida, Gali and Gertler (1998) include the German interest rate
as an important determinant of interest rates in the other ERM countries. But
provided that German interest rate movements can be well explained in terms of
a Bundesbank policy reaction function with respect to German in¯ation and
German output growth, these two factors will also explain some proportion of
interest rate movements in France. Figures 4.1a±4.1c show that German and
French interest rates are correlated, and Figures 4.1d±4.2a show that the same is
true for both in¯ation and output growth, but less so for money growth in Figure
4.2b. From the data's point of view it therefore seems to be unclear whether
interest rate smoothing rules in the ERM member countries like France are best
estimated on domestic or German data or any combination of the two. This is
why we aim at studying the degree of interest rate interaction between European
countries rather than looking at these countries in isolation.
The remainder of the chapter is organised as follows: Section 2 brie¯y
surveys the existing evidence about interest rate smoothing rules for Europe
while Section 3 deals with the ability of interest rate smoothing policies to
achieve the ®nal objective of monetary policy, which is low in¯ation. To get a
handle on this we use a bivariate VAR with in¯ation and interest rates.
Orphanides (1999) refers to this model as a simple prudent policy rule. We
study the short-run and long-run interaction of both variables in this prudent
rule under the very ¯exible identi®cation scheme proposed by King and
Watson (1997). Section 4 deals with the interaction of German and other
European interest rates, and we will examine the degree of long-run and short-
run interest rate co-movements. Section 5 concludes.
Gathering all constant terms in (4.1) results in the typical Taylor rule in which
nominal interest rates
it respond to output
yt and in¯ation
t :
Germany
8
4
a
0
–4
France
–8
France
–12
–16
1971 75 79 83 87 91 95 99
b
10
6
2
–2
–6
1971 75 79 83 87 91 95 99
whereby the in¯ation response coef®cient has a numerical value of 1.5, and
hence real interest rates rise in response to in¯ation. Clarida, Gali and Gertler
(1997, 1998) and Rudebusch and Svensson (1998) introduce a partial
adjustment mechanism for interest rates owing to adjustment costs and
derive the dynamic interest rate smoothing rule:
whereby 0 < < 1 holds for the partial adjustment coef®cient. The estimates
obtained by these authors are reproduced in Table 4.1. The key point about
Table 4.1 is that the long-run in¯ation response coef®cient is larger than unity
for Germany, Japan and the United States, but not for the United Kingdom,
France and Italy. In Clarida, Gali and Gertler (1998) the same result also applies
in many of the auxiliary regressions which include other potential policy
target variables. Rudebusch and Svensson (1998) use a modi®ed interest rate
smoothing rule in their optimal policy simulations, which are based on:
In contrast to the simple Taylor rule their optimal rule involves a long-run
in¯ation coef®cient of larger than 2, whilst the long-run output coef®cient in
the optimal rule is greater than 1.4 Kuttner and Posen (1998), on the other
hand, estimate the interest rate rule in ®rst differences. One of the rules they
estimate is the simple prudent rule:
Table 4.1 Selected coef®cient estimates of Clarida, Gali and Gertler (1998), baseline
regressions
Lagged AR (p)
Country Period In¯ation Output interest rate Constant model
Germany 79.03±93.12 1.31 (0.09) 0.25 (0.04) 0.91 (0.01) 3.14 (0.28) AR(1)
Japan 79.04±94.12 2.04 (0.19) 0.08 (0.03) 0.93 (0.01) 1.21 (0.44) AR(1)
USA 79.10±94.12 1.79 (0.18) 0.07 (0.06) 0.92 (0.03) 0.36 (0.85) AR(2)
USA 82.10±94.12 1.83 (0.45) 0.56 (0.16) 0.97 (0.03) ±0.1 (1.54) AR(2)
UK 79.06±90.10 0.98 (0.09) 0.19 (0.04) 0.92 (0.01) 5.76 (0.69) AR(1)
France 83.05±89.12 1.13 (0.07) 0.88 (0.10) 0.95 (0.01) 5.75 (0.28) AR(1)
Italy 81.06±89.12 0.90 (0.04) 0.22 (0.08) 0.95 (0.01) 7.14 (0.37) AR(1)
Coef®cient
range 0.90±2.04 0.08±0.88 0.92±0.97 ±0.1±7.14
Simple
Taylor rule 1.5 0.5
Table 4.2 Selected in¯ation coef®cient estimates of Kuttner and Posen (1998)
which corresponds closely to the type of interest rate policy rule we will
analyse below. The corresponding -estimates are summarised in Table 4.2.
Note that the in¯ation response estimates are typically larger than unity and
that in¯ation is assumed to be exogenous in Kuttner and Posen (1998). In the
analysis below we will see the potential impact of the possible endogeneity of
in¯ation on these coef®cient estimates.
Let us now turn to this endogeneity issue. Our key argument is that estimating
policy reaction functions is uninformative owing to the problem of
observational equivalence. Take the link between interest rates and in¯ation,
with the latter typically being considered as the monetary policy objective:
without detailed knowledge about the time series properties of the in¯ation
process, reduced form econometric methods are unable to discriminate
empirically between an interest rate policy reaction function (an anti-in¯ation
feedback rule) on the one side and the in¯ationary consequences of the
interest rate policy on the other side. However, as McCallum (1984) has
shown, such tests can be constructed using cross-equation restrictions in a
bivariate vector-autoregressive (VAR) approach.
Like much of the recent literature, we study a VAR representation of policy
and the economy, and we address an important criticism of VAR methods
concerning the possible non-robustness of identi®cation. To be precise, we
aim at relaxing the parameter restrictions required for identi®cation. In
particular, instead of focusing attention on the implications of a single set of
point estimates, we consider a sequence of identi®cation schemes satisfying
economically motivated inequality constraints. This estimation technique is
in the spirit of robustness checks exempli®ed by the studies of Blanchard
(1989), King and Watson (1997) and Bernanke and Mihov (1998). We limit
ourselves to the analysis of bivariate VARs, since the nature of this simple
structural model allows us to relate the long-run comovements of interest rates
and in¯ation to interpretable parameters describing the short-run dynamics of
both series in a bivariate setting. In particular, we specify the space of
Asymmetric Interest Rate Policy In Europe 139
The present chapter follows Geweke (1986), Stock and Watson (1988),
Fisher and Seater (1993), Weber (1994) and King and Watson (1997) and bases
inference about long-run economic propositions on explicit tests of coef®cient
restrictions in bivariate vector-autoregressive models.5 But in order to be able
to test for long-run neutrality (homogeneity) it has been shown that
meaningful tests can be constructed only if both monetary and real variables
satisfy certain non-stationarity conditions, which are spelled out in detail in
Fisher and Seater (1993) and King and Watson (1997). These studies
demonstrate that straightforward neutrality tests, such as imposing the
restriction that the coef®cients of current and lagged monetary impulses in
a regression on real economic variables sum to zero, can be conducted only if
the order of integration of both series is at least one and equal for both series.6
In addition, both series should not be co-integrated.7 Fisher and Seater (1993)
further show that much of the evidence in the older literature on long-run
neutrality or homogeneity violates these non-stationarity requirements, and
hence has to be disregarded.8 Before we present any structural estimates it is
therefore key to discuss brie¯y the unit root and cointegration properties of
the data.
®nding of unit roots in in¯ation and interest rates is consistent with previous
non-stationarity ®ndings in Gali (1992), Weber (1994) and King and Watson
(1997), who also ®nd indications of unit-roots in US in¯ation and interest data
during the postwar period and use both variables in ®rst difference form in
their VARs.
Notes:
Column 3 reports the augmented Dickey±Fuller tests for detrended data or demeaned data,
respectively. Their signi®cance levels are taken from Table 8.5.2. of Fuller (1976), p. 373.
A rejection of the null hypothesis of a unit root at the 1 per cent signi®cance level is marked with **,
Stock's (1991) 95 per cent con®dence intervals for the largest unit root were calculated from the
ADF statistics using Stock's Tables A1 and A2 and the proceedure described in Appendix B of his
(1991) paper.
In addition to the con®dence belts for the estimated roots are displayed.
All ADF statistics are based on regressions including six lagged differences of the variable.
X
p
j
X
p
t i it i it�j j t�j "m
t
4:6
j1 j1
X
p
j
X
p
j
it i t ii it�j i t�j "t
4:7
j1 j1
where i and i represent the contemporaneous effect of interest rate policy
on in¯ation and the contemporaneous response of interest rate policy to
in¯ation, respectively. A more convenient representation of this bivariate VAR
system is:
whereby
Pp j Pp j Pp j
ii
L 1 � j1 ii Lj ; i
L i j
j1 i L ;
L 1 �
j
j1 L as well
Pp j j
as i
L i j1 i L applies. In stacked form this may be re-written as:
" #
j j
i
and �
j
j j ; j 1; 2; :::; p
i ii
In the above notation the long-run multipliers are
i and
i; where
i i
1=ii
1 measures the long-run response of nominal interest rates
to a one unit permanent increase in in¯ation, while
i i
1=
1
measures the long-run response of in¯ation to a permanent unit increase in
interest rates. The long-run Fisher effect implies the restriction
i 1.
As noted by King and Watson (1997), (4.10) is econometrically unidenti®ed
and the restrictions implied by the Fisher effect are no longer testable when
in¯ation is endogenous. Thus, even if the hypothesis that "m t and "t are
uncorrelated is maintained, one additional restriction is required in order to
identify the linear simultaneous equation model. Common identifying
assumptions are that in¯ation is exogenous (
i 0) or predetermined
i 0. Alternatively, the long-run Fisher effect with
i 1 may be imposed
in order to identify the system and estimate the remaining parameters. In
principle it is possible to identify the above model by specifying a value of any
one of the four parameters i ; i ;
i or
i and then ®nding the implied
estimates for the other three.
Germany 79.03±98.10 0.31 0.30 0.01 0.49 0.75 0.79 ±1.8,0 0,1.5 ±4 0.18 (0.21) 0.48 (0.26) ±0.50 (0.28)
France 79.03±98.10 0.27 0.62 0.15 0.47 0.77 0.62 ±0.6,0 0.44 ±2.5 0.01 (0.19) 1.24 (0.51) ±0.18 (0.12)
Italy 79.03±98.10 0.31 0.55 0.04 0.71 0.73 0.66 ±0.04 0.26 0.16 ±0.86 (1.02) 1.13 (0.51) ±0.35 (0.18)
Netherlands 79.03±98.10 0.27 0.61 ±0.01 0.40 0.55 0.48 ±0.2 0.86 0.17 ±0.29 (0.32) 2.15 (0.86) ±0.43 (0.18)
Belgium 79.03±98.10 0.32 0.95 0.08 0.39 0.63 0.47 ±0.3,0.8 ±0.29 0.2 ±0.14 (0.21) 0.80 (0.55) ±0.06 (0.07)
Denmark 79.03±98.10 0.45 1.37 0.10 0.49 0.87 0.50 ±0.1,0.2 0.85 n.a ±0.05 (0.15) ±0.11 (0.48) 0.05 (0.06)
UK 79.03±98.10 0.49 1.08 0.17 1.14 0.86 0.86 ±0.2 1.16 n.a. 4.26 (4.02) 2.27 (0.72) ±0.48 (0.22)
USA 79.03±98.10 0.27 0.67 0.16 0.55 0.68 0.49 ±1.1,0 0.84 0.1 ±0.43 (0.40) 1.89 (0.63) ±0.27 (0.14)
Japan 79.03±98.10 0.44 0.29 ±0.01 0.54 0.70 0.88 0.02 0 ± 0.25 (0.22) 0.28 (0.16) ±0.67 (0.38)
Notes:
All results for the second moments are based on VARs with six lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated
The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level.
The point estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are
145
146 Axel A. Weber
Another interesting ®nding of Table 4.4 concerns the link between the long-
run and short-run in¯ation response coef®cients. Both the Taylor rule and the
interest rate smoothing rules of Clarida, Gali and Gertler (1998) have a built-in
long-run Fisher effect as the neutral policy stance under which the central
bank's interest rate policy neither actively stabilises nor accommodates
in¯ation. The Fisher effect is therefore useful as a benchmark case. By
imposing the Fisher effect (
i 1), we obtain estimates of the short-run
in¯ation response coef®cient (i ) which are signi®cantly larger than one for
the Netherlands, the United Kingdom and the United States, while for
Germany, France, Italy, Belgium and Japan they are signi®cantly greater than
zero but not larger than one. For the latter countries a less than proportional
interest rate adjustment to in¯ation appears to be consistent with a long-run
neutral policy stance, whilst for the Netherlands, the United Kingdom and the
United States the data point towards some degree of short-run interest rate
smoothing policies.
Since the short-run in¯ation response coef®cient for Germany is signi®-
cantly smaller than unity, does this imply that the Bundesbank has
accommodated in¯ation? The answer is no. Consider Figure 4.3, which
displays the coef®cient sensitivity analysis graphically. Economically mean-
ingful policy reaction functions clearly imply i 1. Panel b shows that for
Germany even moderately positive in¯ation response coef®cients
(0 < i < 1:5) are compatible with a long-run Fisher effect (
i 1). According
to Panel c the data also indicate that German in¯ation is in the long run
exogenous with respect to interest rates across a wide range of possible
identi®cation schemes, whilst from Panel a it is obvious that
i 1 implies a
strongly negative immediate impact of interest rates changes on in¯ation (i ).
On the other hand, if we assume in¯ation to be predetermined (i 0), this is
consistent with i 1 only for an initial policy response to in¯ation which is
close to zero (�0:1 <i < 0:1).11 In general, Panel d of Figure 4.3 shows that
there exists a `tradeoff' between the coef®cients i ; and i . This allows the
reader to carry out the following `speci®cation test' concerning the long-run
Fisher hypothesis: if the reader believes that the true value of the pair (i ; i )
lies outside the 95 per cent con®dence region, then the model with the long-
run Fisher effect imposed is rejected by the data at the 5 per cent level. We
believe that monetary policy has long and variable outside lags, and therefore
in¯ation will initially only respond moderately to interest rate changes
(i 0). We furthermore believe that money is neutral in the long-run and
that a long-run Fisher effect exists. This together implies that monetary
policy of the Bundesbank is unlikely to be adequately described by an interest
rate smoothing rule, since a long-run unit effect of in¯ation on interest rates
for Germany is compatible with no short-run policy reaction of interest rates
to in¯ation at all. The same prior beliefs would lead us to conclude from
Figure 4.4 that for the United States only a much larger immediate response
of interest rates to in¯ation would be consistent with a long-run neutral
Asymmetric Interest Rate Policy In Europe 147
Figure 4.3 The link between interest rates policy and in¯ation, Germany, 1979:3±
1998:10
monetary policy stance. This empirical result may simply be a re¯ection of the
fact that the United States actually conducted its monetary policy through an
interest rate smoothing rule, whilst the Bundesbank did not.
For the remaining countries the results in Table 4.4 may be summarised as
follows: Overall, our results are quite similar for all countries, and Figure 4.5
demonstrates this for the con®dence ellipses under the null hypothesis of a
long-run Fisher effect. More speci®cally, our ®ndings for most ERM countries
resemble the results obtained for Germany, while the United Kingdom is more
like the United States. With the exception of the United States, the closest
long-run correlations between in¯ation and interest rates are typically found
for the large economies (Germany, United Kingdom, Japan), while smaller
economies display a lower long-run correlation. This may indicate that during
the ERM German interest rates rather than domestic in¯ation may have been
the dominant driving force behind domestic interest rates in many countries.
We will examine this proposition in more detail below.
148 Axel A. Weber
Figure 4.4 The link between interest rates policy and in¯ation,
United States, 1979:3±1998:10
Since January 1999 European interest rate policy was placed under the authority
of the European Central Bank (ECB) and of®cial interest rates in EMU member
countries are, by de®nition, moving in a perfectly symmetric fashion. Prior to the
creation of the monetary union this was not the case, but owing to nominal
convergence there has been an increasing degree of co-movements between
European rates. In this context it has frequently been argued that Germany has
played the role of the anchor country in the ERM, and that interest rate changes
originating in Germany would ultimately cause symmetric changes in the interest
rates of the remaining ERM countries. To examine this proposition, we ®rst
analyse the data properties and then apply our estimation technique.
UK Netherlands
4
3
USA
Italy
2
Germany
λi,π
1
0
Belgium
–1
Denmark
–2
λπ,i
Figure 4.5 The estimated tradeoff between policy response and the policy effects for
interest rate policy and in¯ation, 1979:03±1998:10
Note: The coef®cients ;i and i; represent the contemporaneous effect of interest rate
policy on in¯ation and the contemporaneous in¯ation response of interest rate policy
respectively.
The con®dence epilses indicate the parameter space for the pair
;i ; i; over which
the data do not reject the long-run Fisher-effect at the 5 signi®cance level.
the tests below have to be interpreted with caution. But in addition to non-
stationarity, the absence of cointegration and hence the non-stationarity of
interest rate differentials vis-a-vis Germany is required. Based on augmented
Dickey±Fuller tests we cannot reject the non-stationarity of interest rate
differentials for France, Italy, the United Kingdom and the United States, while
trend-stationary processes are found for Belgium, Denmark and Japan and a
mean-stationary process is identi®ed for the Netherlands. We will discuss our
®ndings in detail only for the ®rst set of countries.
complicated by two facts. First, the interest rate policy may be transmitted
with lags, and in practice it therefore makes sense to distinguish between the
immediate and the long-run transmission of interest rate policy by using the
dynamic speci®cation:
X
p
j
X
p
j
it i i it i i it�j i i it�j "t
4:12
j1 j1
X
p
j
X
p
j
it i i it i i it�j i i it�j "m
t
4:13
j1 j1
X
p
j
X
p
j
it i i it ii it�j i i it�j "t
4:14
j1 j1
parameters ii ; i i ;
i i or
ii and then ®nding the implied estimates for the
other three parameters.
France 79.03±98.10 0.31 0.59 0.18 0.58 0.87 0.81 ±0.2,0.2 ±0.5,0.9 ±0.3 0.21 (0.14) 0.20 (0.31) 0.04 (0.09)
Italy 79.03±98.10 0.31 0.55 0.12 0.55 0.79 0.46 ±0.4,0.1 0,1.3 0.1 ±0.29 (0.24) 0.56 (0.29) ±0.11 (0.11)
Netherlands 79.03±98.10 0.30 0.56 0.35 0.52 0.56 0.75 ±0.2,0.1 0.1,1.2 0.2 ±0.48 (0.53) 0.71 (0.25) ±0.02 (0.09)
Belgium 79.03±98.10 0.31 0.91 ±0.01 0.53 0.64 0.59 ±0.3,0.0 0.2,2.8 0.1 ±0.39 (0.33) 1.12 (0.54) ±0.13 (0.06)
Denmark 79.03±98.10 0.31 1.44 0.13 0.55 1.33 0.53 ±0.1,0.1 ±1.0,1.8 ± 0.1 0.06 (0.09) 0.53 (0.73) 0.01 (0.04)
UK 79.03±98.10 0.31 1.15 0.11 0.54 0.83 0.31 ±0.3,0.0 0.9,3.6 0.0 ±0.27 (0.20) 1.91 (0.64) ±0.12 (0.06)
USA 79.03±98.10 0.29 0.70 0.04 0.57 0.87 0.54 ±0.8,±0.1 0.8,4.0 0.2 ±0.12 (0.20) 1.82 (0.64) ±0.30 (0.12)
Japan 79.03±98.10 0.31 0.31 0.20 0.55 0.71 0.37 ±0.2 0.4 0.6 ±0.44 (0.36) 0.80 (0.23) ±0.74 (0.35)
Notes:
All results for the second moments are based on VARs with three lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated
VAR (the spectral density matrix of the variables at frequency zero).
The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level. The point
estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are reported in
columns (12)±(14).
Asymmetric Interest Rate Policy In Europe 153
EMU. Let me focus on the ERM countries here and disregard the free ¯oaters,
because for them this is an inadequate restriction.12 But even among the ERM
countries this restriction is inconsistent with a predetermined or a long-run
exogenous German interest rates policy. Except for France the estimated
coef®cients in columns (12) and (14) are predominantly negative and
frequently signi®cant. For the data not to reject the imposed restriction, the
Bundesbank would have had to increase German interest rates in response to
falling interest rates in the rest of Europe. During German uni®cation this was
in part the case, but the restrictive German monetary policy was motivated
exclusively by domestic objectives related to the tax versus credit ®nancing of
uni®cation and not in response to European developments. Furthermore, the
above restriction would imply a relatively high immediate impact of German
interest rate policy on interest rates abroad, ranging from 0.2 for France to 1.12
for Belgium. Except for France this by far exceeds the short-run correlations in
the data and we therefore reject this restriction.
When comparing Tables 4.4 and 4.5, another interesting result becomes
obvious. For the smaller ERM countries, such as the Netherlands, Belgium or
Denmark, the long-run correlations of domestic interest rates with German
interest rates are actually higher than the corresponding correlations with
domestic in¯ation. In this case the obvious interest rate policy rule appears to
be that these countries have predominantly followed interest rate movements
in Germany. But our estimates in Table 4.5 show that whilst interest rates
appear to be interconnected, there is no simple long-run homogeneous
pattern of co-movements during the ERM period.
and Watson (1997), the identi®cation problem becomes much more dif®cult
in this case, since the number of identifying parameter restrictions increases
with the square of the variables in the model. This may pose computational
problems, given that each identifying restriction is iterated hundreds of times
over a wide range of plausible values.
The chapter uses monthly data. The time series and data sources used were:
. Interest rates: Money market rates, International Monetary Fund (IMF),
International Financial Statistics, various issues, line 60b:
± Discount rates, IMF, International Financial Statistics, various issues, line 60.
± Treasury bill rates, IMF, International Financial Statistics, various issues,
line 60c.
± Government bond rates, IMF, International Financial Statistics, various
issues, line 61.
. Consumer prices: IMF, International Financial Statistics, various issues, line
64.
. Money stock (M1): IMF, International Financial Statistics, various issues, line
34.
. Output, industrial production: IMF, International Financial Statistics, various
issues, (line 66b)
Notes
1. Gerlach and Smets (1998) apply such a model to aggregate data constructed for the
Euro area, whilst the other studies focus on various countries separately.
2. Clarida, Gali and Gertler (1998) try to control for the endogeneity of expected
in¯ation by using past interest and in¯ation rates as well as other variables (output,
commodity prices and real exchange rates) as instruments in a two-step nonlinear
two-stage least squares estimator. For details see n. 11 of their paper.
3. Obviously, our choice to employ a bivariate VAR instead of a multivariate VAR
approach involves the risk of a potential misspeci®cation. However, the alternative
approach also suffers from a potential spurious or unbalanced regression problem
since output is typically not found to be a trendstationary process, as proposed by
Clarida, Gali and Gertler (1998). We ®nd that using output growth instead results
in at the best only marginally signi®cant coef®cient estimates. Thus, under a more
appropriate speci®cation the potential misspeci®cation problem largely reduces. In
our view this justi®es the use of a bivariate VAR, but we agree that extending our
model to a multivariate VAR is the obvious next step for future research.
4. See Rudebusch and Svensson (1998), p. 25.
5. Long-run neutrality here implies a zero-restriction on the sum of coef®cients of the
contemporaneous and lagged in¯ation variables in a regression on real interest
rates. Tests of short-run neutrality, such as those conducted in Sims (1972), on the
other hand, impose zero-restrictions on the individual coef®cients of the monetary
impulses. Long-run neutrality is thus a weaker test, and short-run non-neutralities
may well be compatible with long-run neutrality.
156 Axel A. Weber
6. In the present context unit root tests are important since they indicate those cases
for which the neutrality results may be a statistical artefact. This has been shown to
be the case if the two variables under study do not have the same relative order of
integration. There is no implication whatsoever with respect to the absolute order
of integration in the data, given that it is at least one. The absolute order of
integration does determine the appropriate minimum degree of differencing of the
data before conducting neutrality tests, but overdifferencing does not affect the
neutrality propositions. See Fisher and Seater (1993) for details.
7. If it and t are non-stationary and cointegrated, then a ®nite VAR in ®rst differences
does not exist because there would be a unit root in a moving average polynomial,
which is not invertable. Fisher and Seater (1993) brie¯y discuss cointegration in the
context of long-run neutrality restrictions in an Appendix. They show that
cointegration per se does not affect the long-run neutrality restrictions derived.
However, cointegration typically is suf®cient for rejecting long-run neutrality.
8. Long-run neutrality test results will be presented even if unit root tests suggest that
they may be uninformative. The key point is that the unit root tests may indicate in
which case the neutrality results are a statistical artifact.
9. See Clarida, Gali and Gertler (1998), n. 9.
10. Similar unit root tests for output (not reported here) show that in none of the
countries considered here is output (industrial production) driven by a trend-
stationary process, as is assumed in Clarida, Gali and Gertler (1998) for their
measure of the output gap. In line with much of the literature we ®nd output to be
an I(1) process.
11. Note that the corresponding estimates of i in Clarida, Gali and Gertler (1998) for
Germany are also very low (i 0:118) owing to the large AR(1) coef®cient in the
interest rate equation.
12. For the free-¯oating countries the data are clearly inconsistent with this restriction:
in most cases the contemporaneous effect of German interest rate changes on
foreign interest rates would be relatively high, taking minimum values of 0.4 for
Japan to 0.8 for the United States and 0.9 for the United Kingdom. This by far
exceeds the short-run correlations in the data. German interest rates would also be
endogenous and react strongly to interest rate settings abroad.
Asymmetric Interest Rate Policy In Europe 157
References
Bernanke, B. and I. Mihov (1998), `The Liquidity Effect and Long-run Neutrality',
National Bureau for Economic Research Discussion Paper, 6608.
Bernanke, B. and M. Woodford, (1996), `In¯ation Forecasts and Monetary Policy',
Princeton University, October, mimeo.
Blanchard, O., (1989), `A Traditional Interpretation of Macroeconomic Fluctuations',
American Economic Review, 79, 1146±64.
Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in
C.Romer and D. Romer (eds), Reducing In¯ation, Chicago, Chicago University Press.
Clarida, R., J. Gali and M. Gertler (1997) `Monetary Policy Rules and Macroeconomic
Stability: Theories and Some Evidence', New York University, May, mimeo.
ÐÐÐÐ (1998) `Monetary Policy Rules in Practice', European Economic Review, 42, 1033±
68.
Cohen, D. and C. Wyplosz, (1989) `The European Monetary Union: An Agnostic
Evaluation', Centre for Economic Policy Research Discussion Paper, 306.
DeGrauwe, P. (1989) `Is the European Monetary System a DM-Zone?', Centre for European
Policy Studies Working Document, 39.
Duck, N. W. (1988) `Money, Output and Prices: An Empirical Study Using Long-term
Cross Country Data' European Economic Review, 32, 1603±19.
ÐÐÐÐ (1993) `Some International Evidence on the Quantity Theory of Money',
Journal of Money, Credit and Banking, 25, 1±12.
Fisher, M. E. and Seater, J. J. (1993) `Long-Run Neutrality and Superneutrality in an
ARIMA Framework', American Economic Review, 83, 402±15.
Fuller, W. A. (1976) Introduction to Statistical Time Series, New York: John Wiley.
Fratianni, M. and J. von Hagen (1990a) `German Dominance in the EMS: Evidence from
Interest Rates', Journal of International Money and Finance, Vol. 9, pp. 358±375
ÐÐÐÐ (1990b) `The European Monetary System: Ten Years After', in A.H.Meltzer and
C. Plosser (eds), Carnegie±Rochester Conference Series on Public Policy, 32.
Gali, J. (1992) `How Well Does the IS-LM Model Fit the Postwar US Data?', Quarterly
Journal of Economics, may, 709Ð38.
Gerlach, S. and F. Smets (1998) `Output Gaps and Monetary Policy in the EMU Area',
Bank for International Settlements, September, mimeo.
Geweke, J. (1982) `Measurement of Linear Dependence and Feedback Between Multiple
Time Series', Journal of the American Statistical Association, 77, 304±13.
ÐÐÐÐ (1986) `The Superneutrality of Money in the United States: An Interpretation of
the Evidence', Econometrica 54, 1±21.
King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality' Federal Reserve Bank
of Richmond, Economic Quarterly, 83, 69±101.
Kuttner, K. N. and A. S. Posen (1999) `Does Talk Matter After All? In¯ation Targeting and
Central Bank Behaviour', mimeo. CFS Working paper 99/04.
Lothian, J. R. (1985) `Equilibrium Relationships Between Money and Other Economic
Variables' American Economic Review, 75, 828±35.
Lucas, R. E. Jr, (1972) `Econometric Testing of the Natural Rate Hypothesis', In O.
Eckstein (ed.), `The Economics of Price Determination', Washington, DC, Board of
Governors of the Federal Reserve System.
ÐÐÐÐ (1976) `Econometric Policy Evaluation: A Critique. The Phillips Curve and
Labour Markets', Carnegie±Rochester Conference Series on Public Policy, 1, K. Brunner and
A. H. Meltzer (eds), Amsterdam: North-Holland.
158 Axel A. Weber
ÐÐÐÐ (1980) `Two Illustrations of the Quantity Theory of Money', American Economic
Review, 70, 1005±14.
McCallum, B. T. (1984) `On Low-frequency Estimates of Long-run Relationships in
Macroeconomics', Journal of Monetary Economics, 14, 3±14.
Mills, T. C. (1982) `Signal Extraction and Two Illustrations of the Quantity Theory',
American Economic Review, 72, 1162±8.
Mishkin, F. S. (1984) `The Real Interest Rate: A Multi-country Empirical Study', Canadian
Journal of Economics, 1, 283±311.
Orphanides, A. (1999) `The Quest for Prosperity Without In¯ation', Board of Governors
of the Federal Reserve System, May, mimeo.
Rudebusch, G. and L. Svensson (1998) `Policy Rules for In¯ation Targeting', Centre for
Economic Policy Research Discussion Paper, 1999, October.
Sargent, T. J. (1973) `Interest Rates and Prices in the Long Run: A Study of the Gibson
Paradox', Journal of Money, Credit and Banking, 5, 383±449.
Sims, C. A. (1972) Money, Income and Causality', Papers and Proceedings of the American
Economic Association, 62, 540±52.
Stock, J. H. (1991) `Con®dence Intervals for the Largest Unit Root in US Macroeconomic
Time Series', Journal of Monetary Economics, 28, 435±59.
Summers, L. H. (1983) `The Non-adjustment of Nominal Interest Rates: A Study of the
Fisher Effect in J.Tobin (ed.) Macroeconomics, Prices and Quantities: Essays in Memory of
Arthur Okun, Washington DC, Brookings Institution.
Taylor, J. B. (1993) `Discretion versus Rules in Monetary Policy Making', in A.H.Meltzer,
and Plosser, (eds), Carnegie Rochester Conference Series on Public Policy, 39, December,
195±214.
Weber, A. A. (1990) `European Economic and Monetary Union and Asymmetries and
Adjustment Problems in the European Monetary System: Some Empirical Evidence',
Centre for Economic Policy Research Discussion Paper 448.
ÐÐÐÐ (1991) `Credibility, Reputation and the European Monetary System', Economic
Policy, 12, 57±102.
ÐÐÐÐ (1994) `Testing Long-run Neutrality: Empirical Evidence from the G7-
Countries with Special Emphasis on Germany', in A. H. Meltzer and Plosser (eds),
Carnegie±Rochester Conference Series on Public Policy, 41, 67±117.
Wyplosz, C. (1991) `EMS Puzzles', Revista Espanola de Economia, 7(1), 33±66.
Discussion
Carlo A. Favero
159
160 Carlo A. Favero
Identi®cation and speci®cation problems for monetary policy rules are best
understood by brie¯y outlining their derivation. I shall consider here one of
the simplest possible models, taken by Svensson (1998). The central bank faces
the following intertemporal problem:
X
1
Et i Lti
i0
1
L
t � 2
D 4:1
2
where Et denotes expectations conditional upon the information set available
at time t , is the relevant discount factor, t is in¯ation at time t and is the
target level of in¯ation. The central bank is de®ned as a strict in¯ation targeter.
As the monetary instrument is the policy rate, it , the structure of the economy
must be described to obtain an explicit form for the policy rule. We consider
the following speci®cation for aggregate supply and demand in a closed
economy:
where x represents deviations of output from its natural level and is the rate
of in¯ation; zt is a vector of exogenous variables.
The following interest rate rule can be derived from the ®rst-order
conditions for optimality:
1 x r
it r
Et t1 �
D 4:4
x r
x z z
xt zt Et zt1
r r x r
The two issues raised by Weber in Chapter 4 are easily understood within the
framework proposed in Section 1. The identi®cation problem can be referred
to the estimation of the monetary policy rule as a single equation, when the
data are generated by a multi-equation model. A related issue is the validity of
the block-recursivity assumption. In general, it is absolutely true that the
estimation of the full model is much more informative than a single-equation
speci®cation of the monetary policy rule. In fact, tests of the compatibility of
actual behaviour of central banks with different type of in¯ation targeting
models can be conducted only by checking cross-equation restrictions on
parameters of multi-equation models.
However, the two-equation speci®cation for in¯ation and interest rates
adopted in Chapter 4 is also a misrepresentation of the full model, in that it
leaves at least one equation out of the picture. Therefore, the validity of the
Weber critiques on identi®cation might be very limited if the omission of the
equation for output leads to invalid estimation and inference in the two-
equation model.
The point of misspeci®cation as a consequence of interdependence of
European interest rates is easily interpreted as a misspeci®cation of the vector
zt , in the case where the German interest rates are exogenous. If this does not
happen, then one more dimension has to be added to the problem by
considering in¯ation targeting in open economies. As a consequence, the full
model of the economy includes at least four equations, and misspeci®cation
problems induced by considering just two of them are likely to become more
severe.
The two representation are absolute identical (same residuals). The cointegrat-
ing properties of the system suggests the presence of two types of shocks: a
permanent one (to be related to the single common trend shared by the two
variables) and a transitory one (to be related to the cointegrating relation). It
seems therefore natural to identify one shock as permanent, the other as
transitory. Given that we have a stationary system, the identi®cation of shocks
is obtained by deriving long-run responses of the variables of interest to
relevant shocks. From
D4:6 we have:
�1 1 1� L 0 11 L 0
yt � xt
�
0 1 0 1 21 L 0 xt
b11 b12 v1t
D 4:7
b21 b22 v2t
Discussion 163
!
�b11 b21
yt � xt 11 �21 -b12 �b22
11 �21
v1t
�21 b11 11 b21 -21 b12 11 b22
D 4:9
xt 11 �21 11 �21
v2 t
�21 b12 11 b22 0
Note
1. Interestingly there is one case in which such structural identi®cation is achieved
thorugh the standard Choleski decomposition: this is the case in which yt is
weakly exogenous for the estimation of b21 :
References
Clarida R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice:some
International Evidence', European Economic Review, 42, 1033±68.
Svensson L. E. O. (1998) `Monetary Issues for the ESCB', paper presented at the Carnegie±
Rochester Conference; available at < http://www.iies.su.se/leosven/ >.
King R. G. and M. W. Watson (1997) `Testing Long-run Neutrality', Federal Reserve Bank
of Richmond Economic Quarterly, 83, 69±101.
Discussion
Philippe Moutot
Over the past few years, the estimation of policy reaction functions has
become quite popular. Nevertheless, studying empirical literature often seems
much like a `®shing expedition'.2 The Weber's study in Chapter 4 adds some
interesting results to an important issue, namely the issue of robustness.
The chapter argues that, on the one hand, it is not possible to disentangle
the determinants and consequences of monetary policy without prior
restrictions, a fact which is well known in econometrics as the `identi®cation
problem'. On the other hand, it is stated that it is impossible to analyse the
interest rate policy rules of the European countries in isolation, owing to the
interest rate linkages via the ERM. Both problems are not new and ± as far as I
can see ± unsolved in literature so far.
My comments will be threefold. I will ®rst argue that the kind of issues
tackled in this chapter and the techniques used for this purpose are
increasingly relevant, at least from the point of view of a central banker. I
will then discuss the results of the chapter to see what lessons might be learned
from it with regard to the identi®cation of rules, submitting that this chapter
should only be the ®rst step of a long-term research project. I would, however,
caution against the ultimate hope that rules could fully replace discretion in
the context of central banking.
1 Both the issues raised and the techniques used are relevant
The issues
In today's empirical literature, VAR and Vector Error Correction (VEC) models
have become increasingly popular tools. Unfortunately, and as is also well
known, such reduced form models per se do not provide us with any
knowledge about causality.
Following this line of reasoning, traditional estimates of the monetary policy
reaction ± e.g. those referred to as `Taylor Rules' ± cannot in most cases discriminate
between the reaction function ± e.g. monetary policy changes in response to in¯ation,
and the transmission process running from the interest rate to in¯ation, owing
164
Discussion 165
The techniques
One way of dealing with the issue is to apply the Choleski decomposition
which is widely used in the older literature and which requires an ordering of
the variables. Owing to its rather mechanical nature, this procedure is
increasingly being viewed as unsatisfactory.
Another possibility for overcoming this problem is to be found in a
structural approach ± i.e. in imposing restrictions either on the short-run or on
the long-run coef®cients in the variance±covariance matrix.3 Inference can
then be carried out in two ways: one way could be to choose an identi®cation
166 Philippe Moutot
scheme (in fact re¯ecting the researcher's a priori economic assumption about
the nature of the shocks and therefore also about causality) and subsequently
to derive the corresponding point estimates. Another way could be to impose
several identi®cation schemes and to deduce whether the values of the
parameters of interests are in line with them. The second alternative has been
used in the well-known study by King and Watson (1997),4 that provides a
rather ¯exible and elegant scheme for imposing and testing identifying
restrictions and which incidentally also coincides with the spirit of Edward
Leamer's early work.5
This is also the alternative chosen by Weber. In the end, Weber estimates its
parameters by imposing a restriction that is compatible both with economic
theory and with the data. Given the generally low power of the stationarity
tests used by Weber to test for co-integration in small samples, the use of VEC
rather than VAR models might have been considered.
identi®ed with full precision. For each assumption on the short-term impact of
in¯ation on interest rates, there is a corresponding separate estimate of the
interest rate rule.
Second, the hypothesis of a rule combined with a well functioning
economic structure cannot always be con®rmed. This is obviously the case with
the United Kingdom where some more detailed analysis needs to be carried
out. One may wonder if the time series is not too long in that case, mixing too
many different monetary regimes and therefore being misleading. Somehow,
con®rming a rule also depends on the actual period on which it can be
observed.
Third ± and this is meant to be a more philosophical conclusion ± it is not at
all sure whether it is possible for a new central bank like the ECB to ®nd optimal
rules `ex ante'. Owing to this issue of `observational equivalence', some
`learning by doing' is unavoidable. Credibility is of the essence in such a
process. And the way credibility in¯uences the functioning of the economy is
likely to depend on the nature of the rule itself. This is a point which becomes
even more clearly visible when considering the second type of issue tackled by
Weber.
Notes
1. Prepared by Philippe Moutot, with the help of Dieter Gerdesmeier.
2. Shiller (1990), p. 667. Shiller was actually referring to the estimation of the
determinants of the term structure, but his statement seems to be valid for our
considerations as well.
3. See Sims (1986) and Bernanke (1986) for example.
4. King and Watson have also used the framework for the estimation of the long-run
Phillips Curve (i.e. the relation between in¯ation and unemployment) which is of
particular interest for a central bank.
5. See Leamer (1978) and (1983) for a more detailed overview.
6. It should be stressed that a simple correlation between the residuals in the simple
VAR can easily be calculated. However this does not say anything about the
causality. This question cannot be answered without imposing identifying
restrictions and testing for them.
References
Bernanke, B. (1986) `Alternative Explanations of the Money-Income Correlation',
Carnegie±Rochester Conference Series on Public Policy, 25, 49±99.
Goodfriend, M. S. (1991) `Interest Rates and the Conduct of Monetary Policy', (Carnegie±
Rochester Conference Series on Public Policy, 34, 7±30.
King, R. G. and M. W. Watson (1997) `Testing Long-run Neutrality,' Federal Reserve Bank
of Richmond Quarterly, 83(3), 69±101.
Leamer, E. E. (1978) Speci®cation Searches: Ad hoc Inference with Non-experimental Data,
New York: John Wiley.
ÐÐÐÐ`Let's Take the Con out of Econometrics', American Economic Review, 73, 31±43.
Shiller, R. J. (1990) The Term Structure of Interest Rates, in [Friedman, B. M., and Hahn,
F. H. (eds.), Handbook of Monetary Economics, vol. I, Elsevier Science Publishers
B.V., pp. 627±722., 1990].
Sims, C. (1986) `Are Forecasting Models usable for Policy Analysis?', Federal Reserve Bank
of Minneapolis Quarterly Review, 10, 2±16.
Woodford, M. (1998) `Optimal Monetary Policy Inertia', paper presented at the CFS
conference in Frankfurt, October.
5
1 Introduction
Over the past decade, the countries of central Europe have become more alike
in many ways. As the new members of the European Monetary Union (EMU)
prepared for the birth of the Euro on 1 January 1999, their economic policies
became substantially more uniform. All eleven countries in the new Euro Area
had virtually eliminated in¯ation and taken serious steps toward ®scal
consolidation.2 As their monetary and ®scal policies have adjusted to meet
these common goals, the countries' business cycle ¯uctuations appear to have
become more synchronised as well.3 While this makes the job of the European
System of Central Banks (ESCB) easier, numerous dif®cult challenges remain.
Primary among these is the making of policy in the face of the possibility that
it will have differential impacts across the countries of the Euro Area.
The task facing the ESCB is even more complex than that facing countries with
stable monetary regimes, where the measurement of the national and regional
impact of policy has already proved to be extremely dif®cult. The creation of the
ESCB constitutes a `regime shift' in virtually every sense of the term. The
introduction of the Euro seems sure to prompt adjustments in the economies of
the member countries, and these adjustments will probably alter the relationship
between the actions of the central bank and the real economy. That is, the
monetary transmission mechanism of the countries in the Euro Area will change,
making the job of the new European Central Bank (ECB) even more dif®cult than
it is already. But how quickly will it change, and what will it become?
To answer these questions, we must understand the fundamental determi-
nants of the impact of policy actions on output and in¯ation. For insight into
these determinants, I turn to the modern views of the monetary transmission
mechanism, which assign a central role to ®nancial structure. Kashyap and
Stein (1997) provide a starting point; they focus on the importance of the
banking system and go on to emphasise the distributional effects of monetary
policy changes. The conventional wisdom has always been that some
industries are more sensitive to interest rate changes than others, and so
170
Legal Structure, Financial Structure and Monetary Policy Transmission 171
example, while Dornbusch, Favero and Giavazzi (1998, pp. 48±9) do note the
possibility for EU-wide asymmetries resulting from differences in ®nancial
structure, they assert that `the euro will change the way ®nancial markets
work, inducing corresponding changes in the monetary mechanism. In
addition to pervasive deregulation already under way and innovation, the
introduction of the euro will revolutionize the ®nancial structure of Europe.
Europe will in a short period become more nearly like the USA.' McCauley and
White (1997, p. 17) suggest that there may be an acceleration in the rate at
which securities replace loans on the asset side of bank balance sheets and
commercial paper replaces deposits on the liability side. They point to a
`dramatic potential for assets to be stripped out of the banking system' and for
securities markets to absorb as much as one-third of the corporate loans now
originated in European banks.4 Overall, these commentators are speculating
that the increased liquidity of European ®nancial markets brought about by
monetary union will lead to signi®cant consolidation of banks, with mergers
at both the national and the international level, as well as a direct substitution
of traded equities and bonds for bank loans.
Why should we believe that the European ®nancial structure will quickly be
transformed into one that mirrors the one in the US model? Without an
explanation for the evolution of these countries' national ®nancial structures
that is based on their existing differences, such claims are unconvincing. What
accounts for the variation in the ®nancial intermediation systems across
countries? Traditionally, we look to taxes and regulation for an explanation,
and Dornbusch, Favero and Giavazzi (1998) as well as White (1998) do
mention these. Danthine et al. (1999) identify a number of barriers to change
in national ®nancial structures and note the importance of the historical path
that has brought each country's banks to their current state. Danthine et al.
(p. 45) then go on to assert that `legal differences between EU states, in
particular the lack of some form of ``European corporate law'', also remain
important and constitute an additional factor of market segmentation'. Such
disparities in legal structure can explain important economic patterns, and
they can be maintained for long periods of time, signi®cantly delaying the
harmonisation of national banking systems.5
It is my main contention that the differences in ®nancial structure across
the countries of Europe are a consequence of their dissimilar legal structures.
My argument draws on the work of La Porta et al. (1997, 1998), who focus on
the relationship between legal structures and ®nance. They argue that the
structure of ®nance in a country depends on the rights accorded shareholders
and creditors by the laws of that country, as well as on the degree to which
these laws are enforced. The nature of the laws is, in turn, a product of the legal
tradition on which the civil codes of a country are based. La Porta et al.
establish that the character of a country's ®nancial markets depends on the
country's legal structure. Putting their arguments together with the lending
view of the monetary transmission mechanism leads to the possibility that it is
Legal Structure, Financial Structure and Monetary Policy Transmission 173
Table 5.1 Effectiveness of 14cy and the origins of the legal system
Impact of policy
Index of effectiveness
Legal family of monetary policy On output on in¯ation
Notes:
Index of effectiveness of monetary policy, from Table 5.5, is based on ®nancial structure variables
described in Section 3 of the text, with higher values implying a larger expected impact of interest
rate changes on output and prices.
The impact of policy on output and in¯ation, from Table 5.6, is a measure of the maximum
response, in percentage points, to an interest rate movement of 100 basis points, estimated using a
small-scale structural model.
Countries are classi®ed by the origin, or family, of their legal structure, and group means are reported
based on data for Ireland, the United Kingdom and the United States (English common law);
Denmark and Sweden (Scandinavian common law); Belgium, France, Italy, Portugal, and Spain
(French civil law); and Germany (German civil law).
the legal system in a country that forms the basis for the structure of ®nancial
intermediation and, hence, for the impact of monetary policy on output and
prices.
Table 5.1 reports the empirical ®ndings that support the basic conclusion of
the chapter. After classifying countries by the origin, or `family' of their legal
structure, I calculate for each family the average level of an index of monetary
policy's likely effectiveness (based on banking system size, concentration and
health, with a higher value implying greater effectiveness) and the estimated
impact of an interest rate change on output and in¯ation (from a small-scale
structural model). The results suggest that a country's legal structure, ®nancial
structure and monetary transmission mechanism are interconnected. The
clear pattern is that the predicted effectiveness and its measured impact vary
systematically based on the origin of a country's legal system. Countries with
better legal protection for shareholders and debtors (countries with a legal
structure based on English common law) have ®nancial structures in which
the lending channel of monetary transmission is expected to be less potent;
for these countries, the measured impact of an interest rate change on output
and in¯ation is lower.
The implication is that unless the laws governing shareholder and creditor
rights and the enforcement of those laws are harmonised across the members
of EMU, monetary policy will continue to have a differential impact. Put
slightly differently, it is my belief that the ®nancial structures in the countries
of the Euro Area will not converge into one large US-style system unless there
are dramatic legislative changes. If such legal harmonisation occurs ± that is, if
the civil codes protecting shareholders and creditors are made uniform across
174 Stephen G. Cecchetti
the countries that have entered the monetary union ± then the regional
variation in the impact of interest rate changes on output and in¯ation should
decrease.6 But if legal convergence does not occur, ®nancial structure will
remain heterogeneous, and so will the monetary transmission mechanism,
and the job of the Eurosystem will be to construct an appropriate policy that
takes these asymmetries into account.7
The remainder of this chapter provides the building blocks for this
argument. In Section 2, I provide a brief survey of the theories of the
monetary transmission mechanism, focusing on the importance of ®nancial
structure to an understanding of monetary transmission. Section 3 assesses the
national banking systems, including measures of overall size, concentration,
health and the relative importance of non-bank ®nance. Overall, this analysis
allows me to evaluate the likely strength of the lending channel across
countries. Section 4 reports estimates, for a set of ten countries, of the impact
of an interest rate increase on output and in¯ation. These estimates follow the
pattern that is expected: countries where ®nancial structure data suggest that
the lending channel should be strong exhibit more sensitivity to monetary
policy movements. In Section 5, I present the data and arguments from La
Porta et al. (1997, 1998) on the relationship between legal and ®nancial
structures. This allows me to test the prediction that countries with poor
shareholder and creditor protections and poor law enforcement will have less
developed ®nancial systems and greater sensitivity of output and in¯ation to
interest rate changes. While far from being de®nitive, the results are consistent
with my main hypothesis: differences in legal systems give rise to variations in
national ®nancial structures, and these variations in turn lead to divergences
in monetary transmission mechanisms. So long as the legal systems of the
Euro Area countries remain distinct, the impact of interest rate changes across
these countries will differ.
Eichenbaum and Evans (1997). They distinguish three sets of theories: one set
based on sticky wages, a second based on sticky prices and a third built on the
idea of limited participation. The sticky-wage and sticky-price models, which
are the most familiar, rest on the idea that there are costs to nominal price and
wage changes, and so adjustments are infrequent. In limited participation
models, introduced in Rotemberg (1984), individuals (households) are unable
to adjust their cash balances suf®ciently rapidly in response to changes in the
environment ± that is, households have a limited ability to participate in
®nancial markets, and so must commit themselves to certain portfolio
holdings for relatively long periods of time.8
The sources of nominal rigidities are relatively unimportant for the
discussion of the mechanism by which interest rate changes have short-run
effects on output and prices, and so I will move directly to a discussion of the
current theories of the transmission mechanism.9 Our current views are based
on the work of Bernanke (1983), Bernanke and Blinder (1992) and Bernanke
and Gertler (1989, 1990). These authors distinguish between the traditional
money view, in which interest rate movements affect the level of investment
and exchange rates directly, and the lending view, in which ®nancial
intermediaries play a prominent role in transmitting monetary impulses to
output and prices. I will describe each of these in turn.
The traditional view, which is largely the foundation for the textbook IS ±
LM model, is based on the notion that reductions in the quantity of outside
money raise real rates of return. This outcome has two effects, the ®rst directly
from interest rates to investment and the second through exchange rates. An
interest rate increase reduces investment, as there are fewer pro®table projects
available at higher required rates of return. A policy action induces a
movement along a ®xed marginal-ef®ciency-of-investment schedule. This
interest rate channel will be more powerful the less substitutable outside
money is for other assets. The exchange rate channel is also familiar from
textbook models. Here, an interest rate increase results in a real appreciation of
the domestic currency, reducing the foreign demand for domestically
produced goods. Regardless of whether the transmission mechanism occurs
through the interest rate channel or the exchange rate channel, there is no real
need to discuss banks. In fact, there is no reason to distinguish any of the
`other' assets in investors' portfolios. This is a simple two-asset model.
An important implication of this traditional model of the transmission
mechanism concerns the incidence of the investment decline. Since there are
no externalities or market imperfections, only the least socially productive
projects ± those with the lowest rates of return ± go unfunded. As a result, the
capital stock is marginally lower, but, given that a decline is going to occur, the
allocation of the decline across sectors is socially ef®cient.
As most of the surveys cited earlier emphasise, the lending view has two
parts ± one that focuses on the impact of policy changes on borrower balance
sheets and another that focuses on bank loans. In both, the effectiveness of
176 Stephen G. Cecchetti
policy depends on capital market imperfections that make it easier for some
®rms to obtain ®nancing than others. Information asymmetries and moral
hazard problems, together with bankruptcy laws, mean that the state of a
®rm's balance sheet has implications for its ability to obtain external
®nance.10 By reducing expected future sales and by increasing the cost of
rolling over a given level of nominal debt, policy-induced increases in interest
rates (which are both real and nominal) cause a deterioration in the ®rm's net
worth. Furthermore, there is an asymmetry of information in that borrowers
(®rms) have better information about the potential pro®tability of investment
projects than do creditors (banks). As a result, as the ®rm's net worth declines,
the ®rm becomes less creditworthy because it has an increased incentive to
misrepresent the riskiness of potential projects ± an outcome that will lead
potential lenders to increase the risk premium they require when making a
loan. The asymmetry of information makes internal ®nance of new
investment projects cheaper than external ®nance.
More important for the transmission mechanism per se is that some ®rms are
dependent on banks for ®nance and that monetary policy affects bank loan
supply. A reduction in the quantity of reserves forces a reduction in the level of
deposits, which must be matched by a fall in loans. Nevertheless, lower levels
of bank loans will have an impact on the real economy only insofar as there are
®rms without an alternative source of investment funds.
Substantial empirical evidence supports the importance of both capital
market imperfections and ®rm dependence on bank ®nancing. Kashyap and
Stein (1997) provide a summary of two types of studies. The ®rst type suggests
that banks rely to a large extent on reservable-deposit ®nancing and that, for
this reason, a contraction in reserves will prompt banks to contract their
balance sheets, reducing the supply of loans. The second type establishes that
there are a signi®cant number of bank-dependent ®rms that are unable to
mitigate the shortfall in bank lending with other sources of ®nance. Overall,
recent research does imply the existence of a lending channel.11
Models of monetary policy transmission based on ®nancial structure suggest
a natural place to begin looking for sources of cross-country differences in the
monetary transmission mechanism. The prediction is that, overall, the
transmission mechanism will be stronger in those countries where ®rms are
more bank dependent, and where the banking system is less healthy and less
concentrated. In the ®rst instance, ®rms that have less direct access to capital
markets are unable to blunt the effect of a contraction in bank loans. In the
second, banks themselves have restricted access to non-reservable deposits and
are forced to contract their balance sheets by more for a given change in
policy. In Section 3, I examine data on national ®nancial structure and try to
rank countries based on the likely strength of the transmission mechanism. To
the extent that these cross-country differences are present, then the lending
view implies that they will persist until the ®nancial structures become more
uniform.12
Legal Structure, Financial Structure and Monetary Policy Transmission 177
In assessing the likely impact of an interest rate change on output and prices
in the various countries of the EMU, I follow the work of Kashyap and Stein
(1997) and assemble data on the size and concentration of the banking
systems, along with measures of banking system health, the importance of
bank ®nancing and the size of ®rms. The indicators are chosen to conform as
closely as possible to the economic quantities that the lending view suggests
should be important. The balance sheets of large, healthy banks are not as
sensitive to policy, because reserve contractions can be readily offset with
alternative forms of ®nance that do not attract reserve requirements. In
addition, I examine measures of the development of equity and debt markets
in the EMU countries. Firms with ready capital market access, which are more
likely to be found in countries with extensive secondary securities markets,
will be better insulated from bank loan-supply contractions. Combining these
measures, I construct an index of the probable strength of the monetary
transmission mechanism.13
Table 5.2 Size and concentration of the banking industry, by country, 1996
Other countries
Japan 556 4 30
United States 10 803 40 17
Note: Cobcentration ratios are calculated as the percentage of each country's bank assets accounted
Table 5.3 Measures of banking industry health, by (country, 1996, per cent
Other countries
Japan 0.01 0.75 1.17 1.03 3.32(44)
United States 1.42 0.10 2.68 3.51 1.73 (344)
Notes: Except for the Thomson ratings, all ®gures in the table are calculated as a percentage of total
bank assets. In column (5), the number of banks rated by Thomson in each country and used to
compute the average appears in parentheses.
Sources: See the Data Appendix (p. 190).
countries can be divided into three groups. Austria, Ireland, Italy, Portugal and
Greece appear to have the least well developed external capital markets. They
have small equity and bond markets, and bank loans account for a high
percentage of ®rm ®nancing. By contrast, Belgium, Denmark, Sweden, the
United Kingdom and the United States all have substantial secondary capital
markets, and banks are a less important source of ®nance. The remaining six
countries are somewhere in between these two groups.
Table 5.5 summarises the material in Tables 5.2±5.4 and suggests the overall
relative strength of monetary policy in the fourteen EU countries, Japan and
the United States. The ®nal column, `Predicted effectiveness of monetary
policy,' reports a measure of the effects of monetary policy on output and
in¯ation, where higher values suggest a stronger lending channel and
therefore a larger impact. Overall, the pattern is very similar to the one
reported in Kashyap and Stein (1997, Table 6). Most important, the predicted
effects of interest rate movements vary greatly across countries. For example,
looking at the EMU countries, one would expect that a given interest rate
180 Stephen G. Cecchetti
Other countries
Japan 2;34 18.56 67 39 59
United States 8,479 31.94 111 64 21
Notes: Market capitalisation is the year-end value of ®rms listed on major exchanges. For the United
States, three exchanges are used; for Japan, eight; and for each of the remaining countries, one.
Sources: See the Data Appendix (p. 190).
change would have the most impact in Austria and Italy, countries in which
small banks are relatively important, the banking systems are less healthy and
®rms have little access to non-bank sources of ®nance. The opposite is true of
Belgium, Ireland and the Netherlands, where the banking systems are large
and healthy and non-bank ®nance is readily available; in these countries,
interest rate movements would be expected to have a more muted impact.15
The conclusions of this section could be criticised as applying only to the
pre-EMU period. But will the introduction of the Euro be a catalyst for the
harmonisation of ®nancial structure across the EMU? I take this question up in
more detail later, but at this point I will simply mention that the European
Central Bank (1999) report, Possible Effects of EMU on the EU Banking Systems in
the Medium to Long Term, provides very little evidence to suggest that an
increase in either international banking competition or securitisation and
disintermediation will occur quickly.
Legal Structure, Financial Structure and Monetary Policy Transmission 181
Table 5.5 Summary of factors affecting the strength of the monetary transmissions
mechanism
Availability Predicted
Importance Bank of alternative effectiveness
of small banks health ®nance of monetary policy
Country (1) (2) (3) (4)
Other countries
Japan 2 4 2 2.67
United States 3 1 1 1.67
Notes: Column (1) is based on Table 5.2; column (2), on Table 5.3; and column (3), on Table 5.4.
Column (4) is an average of columns (1), (2) and (3).
Testing the proposition that the banking system's concentration, health and
importance have a material impact on the monetary transmission mechanism
requires an estimate of the effects of an interest rate change on output and
prices. Numerous studies report such estimates for some or all of the countries
of the EU. These include Gerlach and Smets (1995), who estimate a three-
variable structural vector autoregression based on long-run restrictions; de
Bondt (1997), who presents estimates of the impact of policy on output and
prices for Germany, France, Italy, the United Kingdom, Belgium and the
Netherlands that are based on the work of other authors; Dornbusch, Favero
and Giavazzi (1998), who report estimates of the impact of policy on output
and prices derived from both small vector-autoregressive models and large
structural models for Italy, Germany, France, Spain, Sweden and the United
Kingdom; Kieler and Saarenheimo (1998), who study France, Germany and
182 Stephen G. Cecchetti
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Italy Portugal
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Spain Denmark
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Sweden United Kingdom
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0
United States
0.5
0 Output response
–0.5 Inflation response
–1.0
–1.5
2 4 6 8 10 12 14 16 18 20 22 24 26
Figure 5.1 Reaction of output and in¯ation to an interest rate increase of 100 basis
points, quarterly, by country
Sources: Cecchetti (1996); Cecchetti, McConnell and Perez Quiros (1999).
Legal Structure, Financial Structure and Monetary Policy Transmission 183
184
Output In¯ation
Other countries
Japan Ð Ð Ð Ð Ð
United States ±0.07 6 ±0.017 12 ±3.27
Sources: Estimates for the United States are from Cecchetti (1996); those for the remaining countries are from the estimation of Ehrmann's model in
Cecchetti, McConnell and Perez Quiros (1999).
Legal Structure, Financial Structure and Monetary Policy Transmission 185
differences. As noted earlier, La Porta et al. (1997, 1998) have examined the
relationship between a country's legal system and its ®nancial system. The
premise of their work is that investors provide capital to ®rms only if the
investors have the ability to get their money back. For equity holders, this
means that they must be able to vote out directors and managers who do not
pay them. For creditors, this means that they must have the authority to
repossess collateral. In addition to having nominal legal rights, these groups
must also have con®dence that the laws will be enforced.
La Porta et al. (1997, 1998) collect data on the legal systems in forty-nine
countries. They argue that all of these legal systems belong to one of four
families: English common law, French civil law, Scandinavian civil law and
German civil law. With regard to shareholder rights ± speci®cally, the ability of
shareholders to vote directors out ± English common law countries have the
best protections and French civil law countries have the worst. The pattern is
similar for creditor rights, which entail the right to reorganise or liquidate a
®rm. The pattern for enforcement is a bit different: Scandinavian civil law
countries have the most rigorous law enforcement, while French civil law
countries have the most lax.
Other countries
Japan 3 2 8.98 German
United States S 1 10.00 English
Table 5.7 reproduces a portion of Table II from La Porta et al. (1997). The
column labelled `Shareholder rights' reports an index that is higher when
shareholders ®nd it less costly and dif®cult to vote directors out. The column
labelled `Creditor rights' reports an analogous index that is lower when
creditors experience less dif®culty gaining possession of property that has
been used to collateralise a bond or loan. Enforcement is an assessment of
countries' rigour in carrying out their laws, with a higher score implying more
aggressive enforcement. Finally, the table reports the legal family from which
each country's laws are derived.
Using these data to examine the relationship between shareholder rights
creditor rights, and enforcement on the one hand, and the concentration of
ownership and the availability of external ®nance on the other, La Porta et al.
(1997, 1998) come to two conclusions. First, corporate ownership is more
concentrated in countries where shareholders and creditors are poorly
protected by both the substance of the law and its enforcement. Second, and
more germane to the current discussion, countries with weaker legal rules and
less rigorous law enforcement have smaller and narrower capital markets.
Overall, English common law countries have the least concentration of
corporate ownership and the largest and deepest capital markets. French civil
law countries have the most concentrated ownership and the smallest capital
markets. La Porta et al. (1997 p. 1131) conclude that the `differences in the
nature and effectiveness of the ®nancial systems around the world can be
traced, in part, to differences in investor protection against expropriation by
insiders, as re¯ected by legal rules and the quality of their enforcement'. Their
®ndings are con®rmed by the data in Table 5.4, which show clearly that the
United States and the United Kingdom have much more extensive capital
markets than France and Italy.
Following the demonstration in La Porta et al. (1997, 1998) that the systematic
variation in systems of corporate governance and ®nance across countries can
be tied to the differences in the countries' legal systems, I ask if the variation in
the predicted strength of the lending channel and the estimated impact of
interest rate movements on output and in¯ation can be traced to these same
legal differences.20 To address this question, I combine the data from Table 5.5
on the predicted strength of the lending channel of monetary transmission
and from Table 5.6 on the size of the impact of interest rate movements
on output and in¯ation with the measures of cross-country differences in
legal organization from Table 5.7. In Table 5.8, I report the results of two
straightforward exercises. The ®rst separates the countries by the origin of their
legal system and constructs group averages for the effectiveness and impact of
monetary policy from column (4) of Table 5.5 and columns (1) and (3) of Table
5.6 (Table 5.8, top panel). The results follow the pattern predicted by the index
Legal Structure, Financial Structure and Monetary Policy Transmission 187
Table 5.8 Testing the relationship between cross-country differences in legal structure
and monetary policy effectiveness
Group mean
Notes: `Predicted effectiveness' is drawn from column (4) of Table 5.5; the `Impact of policy', from
columns (1) and (3) of Table 5.6. The instrumental variables regression is of columns (1) and (3) of
Table 5.6 on column (4) of Table 5.5, with columns (1), (2) and (3) of Table 5.7 as instruments. All of
the results in this table use only the eleven countries for which there are estimates in Table 5.6:
Ireland, the United Kingdom and the United States (English common law); Denmark and Sweden
(Scandinavian common law); Belgium, France, Italy, Portugal and Spain (French civil law); and
Germany (German civil law).
a
average excludes Belgium.
different from zero. The results for in¯ation are much less satisfactory: the
measures of ®nancial structure appear to be uncorrelated with the impact of
policy on prices. Because of the small size of the sample (eleven countries), the
estimates are all fairly imprecise, and so I treat them as being only suggestive.
7 Concluding remarks
Among the many challenges facing the new Eurosystem is the possibility that
the regions of the Euro Area will respond differently to interest rate changes. In
this chapter, I have suggested that differences in ®nancial structure are a
proximate cause for these national asymmetries in the monetary policy
transmission mechanism. Moreover, I have proposed that these differences in
®nancial structure are likely a result of the EU countries' diverse legal
structures. The evidence, although circumstantial, is consistent with this view.
Most economists believe that the monetary transmission mechanism will vary
systematically across countries with differences in the size, concentration and
health of the banking system, and with differences in the availability of
primary capital market ®nancing. The countries of the EU differ quite
dramatically in all of the dimensions that would seem to matter, leading to the
prediction that the impact of interest rates on output and prices will not be
consistent across countries. While the estimates of the impact of interest rate
changes on output and in¯ation tend to be quite imprecise, they do differ, and
in the way that is predicted by the state of the countries' ®nancial systems.
Finally, we can trace differences in ®nancial structure, the size and scope of
capital markets, and the availability of alternatives to bank ®nancing to
differences in the countries' legal structures.
What does this mean for the future of ®nancial markets and monetary
policy in the Euro Area? Will the European banking system become more like
that of the United States? The arguments presented here suggest that unless
legal structures are harmonised across Europe, ®nancial structures will remain
diverse, and so will monetary transmission mechanisms. It will not be enough
to make regulatory structures more similar, since such a change will not, in
and of itself, alter the structure of capital markets. In other words, I do not view
regulatory competition as a force to eliminate the asymmetries in the ®nancial
intermediation systems of the EU.21 As the ECB (1999) report makes clear, this
force has been very weak in the past and is expected to be weak in the future.
While we may see cross-border mergers and acquisitions of ®nancial sector
®rms that take advantage of the expertise of those already doing business in a
region,22 only a decision to change the existing legal structures so that
shareholders and creditors in all EU countries enjoy the same rights will force
the movement to a US- style ®nancial structure.
Legal Structure, Financial Structure and Monetary Policy Transmission 189
Data Appendix
The data sources for Tables 5.2±5.4 in this chapter are as follows:
Table 5.2
Number of institutions and concentration ratios: For all countries, concentration is
calculated as the assets of the top ®ve banks as a percentage of total bank assets.
Population: International Monetary Fund, International Financial Statistics (January
1999), country report tables, l. 99z, mid-year estimates for all countries.
Austria: Austrian National Bank web pages < http://www.oenb.co.at/stat-monatsheft/
tabellen/2001p.htm >, Ingesamt, Hauptanstalten, for number of institutions; and
< http://www.oenb.co.at/stat-monatsheft/tabellen 2000_5p.htm >, Alle Sektoren, Summe
Aktiva (Ohne Rediskonte), for total assets; Austrian National Bank, Economic Analysis
Division, for assets of top ®ve and top ten banks.
Belgium: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports
on bank balance sheets, p. 36, l. 37 (under supplementary information), for number of
institutions; Bank of Belgium, Financial and Economic Statistics Division, for total assets
of credit institutions and for share of top ®ve banks.
Finland: Bank of Finland, Financial Statistics Desk, for all ®gures.
France: Bank of France, Monetary Research and Statistics Division (DESM-SASM) for all
®gures on credit institutions.
Germany: Deutsche Bundesbank, Monthly Report (May 1998), p. 16, Table IV.1, column
1, for number of institutions; Deutsche Bundesbank, Department of Controlling,
Accounting and Organisation, Division C-2, for share of top ®ve banks.
Ireland: Central Bank of Ireland, Monetary Policy and Statistics, for number and total
assets of all credit institutions (which include licensed banks, building societies, state-
sponsored ®nancial institutions and savings banks); IBCA BankScope database, for assets
of top ®ve banks.
Italy: Bank of Italy, Research Department, for all ®gures.
Netherlands: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 192, l. 37 (under supplementary information), for
number of institutions; De Nederlandsche Bank, Annual Report (1997), Tables 1, 2.1 and
2.2, for assets of top ®ve banks and for total assets of monetary institutions.
Portugal: Bank of Portugal web page < http://www.bportugal.pt/publish/frpu-
blish_e.htm >, Chart VIII.1 and Table VIII.2, for number of institutions and share of
top ®ve banks. OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 231, l. 25, for total assets of commercial banks.
Spain: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on
bank balance sheets, p. 236, l. 37 (under supplementary information), for number of
banks; Bank of Spain, Statistical Bulletin (June 1998), Tables 61.1 (p. 271), 62.1 (p. 281),
63.1 (p. 291), sum of column 1 in all tables, for total assets of banks, savings banks, and
credit cooperatives; IBCA Bankscope database, for assets of top ®ve banks.
Denmark: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 64, l. 37 (under supplementary information), for
number of institutions; Denmark National Bank web page < http://www.nationalban-
ken.dk/nb/nb.nsf/alldocs/ F15D9E8CF275ED1A2412565B4003E8BD5, > for total assets;
IBCA BankScope database, for assets of top ®ve banks.
Greece: Hellenic Bank Association, The Greek Banking System (April 1998), p. 87, for
number of institutions, total assets, and assets of top ®ve banks.
190 Stephen G. Cecchetti
Sweden: Sveriges Riksbank, Statistical Yearbook (1996), p. 17, Table 6, for number of
banks; Sveriges Riksbank, Financial Statistics Department, for share of top ®ve banks.
United Kingdom: British Bankers Association, Annual Abstract of Banking Statistics
(1997), Table 1.04, for number of institutions; Bank of England, MFSD, for shares of top
®ve banks (data relate to all banks and building societies operating in the United
Kingdom and so include the business of foreign-owned af®liates in the United
Kingdom).
Japan: Bank of Japan, International Department, for all ®gures for banks and other
deposit-taking institutions, end of ®scal year 1996 (March 1997).
United States: Federal Financial Institutions Examination Council (FFIEC), Reports of
Condition (call reports database), for all ®gures for commercial banks.
Table 5.3
Bank data: McCauley and White (1997), Table 1. Federal Reserve Bank of New York staff
calculations for Austria, Belgium, Greece, Ireland, Italy and Portugal, based on ranking of
asset size from IBCA BankScope database. In each country, banks were chosen according
to 1997 assets. Return on assets, loan loss provisions, net interest margin and operating
cost are drawn from IBCA BankScope database.
Thomson ratings: Thomson BankWatch database.
Table 5.4
Number of publicly traded ®rms and market capitalisation: International Finance
Corporation, Emerging Stock Markets Factbook (1997), pp. 17 and 23 (also available on
the Wall Street Journal web site < http://update.wsj.com/public/resources/documents/gi-
tab5.htm >).
Population: See sources for population data in Table 5.2.
Privately issued debt: Bank for International Settlements, International Banking and
Financial Market Developments (February 1998), pp. 46±7, Tables 14 and 15, amount
outstanding, December 1996 ®gures; sum of ®gures from Table 14 (international debt
securities) and Table 15 (domestic debt securities).
GDP: International Monetary Fund, International Financial Statistics (January 1999),
country report tables, l. 99b.c for all countries. Year-average exchange rates used for
conversion into US dollars (local currency per US dollar, l. rf for all countries).
Bank loans: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, l. 16 on pp. 27, 35, 63, 67, 91, 115, 143, 159, 163, 167,
191, 231, 235, 251, 259, 263, 303, 307 and 315.
Notes
1. I gratefully acknowledge comments from and discussions with Ignazio Angeloni,
Paul Bennett, Stefan Gerlach, Heinz Hermann, Beverly Hirtle, James Kahn, Anil
Kashyap, Kenneth Kuttner, Gabriel Perez Quiros, Margaret Mary McConnell,
Robert Rich and especially Paolo Pesenti. I thank Michael Ehrmann for providing
his programs, and Rama Seth for helping with the data.
2. Throughout the chapter, I refer to the eleven countries of EMU but provide
information on only ten. Luxembourg is not included.
3. See Angeloni and Dedola (1998).
4. Similar points are made by White (1998), who suggests that competition in
banking may be about to increase in Europe, stimulated by the introduction of the
Legal Structure, Financial Structure and Monetary Policy Transmission 191
Euro. In addition, a ECB (1999) study suggests that EMU may speed up the process
of disintermediation and lead to a more geographically diversi®ed and inter-
nationalised banking system.
5. For example, within the United States, more than 10 per cent of ®rms with assets
exceeding $1 million have chosen to incorporate in Delaware, a state with less than
half of one per cent of the country's population. Why is this? The answer can be
found by considering how the development of Delaware's legal structure has
differed from the development of the legal structure in other states. Originally,
large ®rms were incorporated in New Jersey because the state, in exchange for
incorporation fees and franchise taxes, had liberalised its corporation law to allow
various mergers and cross-holdings that were disallowed elsewhere. State law also
gave very strong power to corporations' directors (Grandy, 1989). Delaware copied
New Jersey's statutes and then bene®ted from changes made to New Jersey's law by
Governor Woodrow Wilson in 1913. As this example suggests, the economic
structure has its source in the legal structure.
6. I should note that ®rms in countries that act slowly will be put at a competitive
disadvantage, and so they might pressure their governments to speed up the
legal changes. The potential strength of such regulatory competition is an open
issue.
7. There is an alternative. A company may move to a country where the ®nancial
system better suits its needs. The La Porta et al. measures, reported in Table 5.7,
suggest that the United Kingdom is the best country in the European Community
in which to issue both bonds and stocks, and so ®rms that wish to have ready
access to primary capital market ®nancing may tend to concentrate there. But for
this strategy to be successful, ®rms would have to reincorporate and move assets
into the alternative jurisdiction. The assets must move to provide the proper
guarantees to investors. All of this seems unlikely.
8. In addition to the differences in the type of nominal rigidity, there are variations in
the way in which the rigidites are modelled. These variations are more than formal;
they have very different implications for the dynamic effects of nominal shocks on
real variables. Different modelling strategies are based on differences in the timing
of price or wage change decisions. There are three basic schemes used, based on
Fischer (1977), Taylor (1980) and Calvo (1983), and they create very different
dynamic responses of real variables to nominal shocks. Fischer, for example,
assumes prices are predetermined, meaning that at some time agents set prices for
some number of future periods: the level of prices set on the decision date can differ
for the different periods before the next decision date. In this model, the impact of
a nominal shock lasts for only as long as it takes for all price setters to have a chance
to reset their price schedules. In the Taylor model, prices or wages are assumed to be
®xed, meaning that their nominal value does not vary between decision dates.
When prices or wages are ®xed, nominal shocks die out only asymptotically. In
Calvo's (1983) model, price-setters change their prices according to a Poisson
process, leading to a variety of possible dynamics.
9. Longer-run considerations, such as the potential costs or bene®ts of modest levels
of in¯ation, critically depend on understanding the sources of nominal rigidity. For
example, Akerlof, Dickens and Perry (1996) and Groshen and Schweitzer (1997)
consider whether small positive levels of aggregate in¯ation can facilitate real
adjustments in the presence of an aversion to nominal wage declines, suggesting
that the long-run goal for in¯ation might be positive. But Feldstein (1996)
contends that the tax distortions created by in¯ation reduce the level of output
192 Stephen G. Cecchetti
permanently, an argument that suggests that the optimal level of in¯ation may
even be negative. Overall, most economists now seem to agree that in¯ation leads
to lower levels of real output and may even retard long-run growth. See Feldstein
(1999) for a summary.
10. As emphasised by Kashyap and Stein (1994), this assertion applies to both ®nancial
and non-®nancial ®rms.
11. This is not to say that the traditional mechanisms, operating through interest rates
and exchange rates, are not present as well. Unfortunately, it has proved to be very
dif®cult to disentangle the individual importance of the various channels of
transmission.
12. It is important to note that there can be signi®cant cyclical and secular changes in
the strength of the lending channel as the health and concentration of the banking
system change, and as capital markets become deeper and broader.
13. After I collected the data for this section, the ECB issued its report Possible Effects of
EMU on the EU Banking Systems in the Medium to Long Term. The Appendix tables in
that report contain much of the same information presented here.
14. Throughout the analysis, I omit Luxembourg.
15. A signi®cant failing of this analysis is the assumption that these relative rankings
are not changing over time. Surely, if I had chosen different dates to measure the
relative health and concentration of countries' banking systems, I would have
created a different set of rankings for the ®rst two indicators. It is entirely possible
that both the relative importance of small banks and the health of the banking
system will become increasingly uniform across countries, leaving only differences
in external ®nance.
16. See Appendix A in Ehrmann (1998) for additional details.
17. The results for the United States are derived from Cecchetti (1996).
18. Although he reports estimates for thirteen countries, the estimates for three of
these countries appear to be dif®cult to interpret. In the case of Finland, for
example, the impact of monetary tightening is to increase output, not decrease it.
For Austria and the Netherlands, we have not been able to replicate the results in
the current version of Ehrmann's paper.
19. Figures 1±13 in Ehrmann (1998) show that the impulse response functions are
rarely signi®cantly different from zero. The same point is made in Cecchetti (1998)
and Cecchetti and Rich (1999).
20 White (1998) makes a related point when he notes that the legal, tax, regulatory
and supervisory frameworks within which ®nancial institutions operate differ
signi®cantly across the various countries of the EU. All of these differences make
direct competition more complex and less appealing. He goes on to focus on
differences in the EU countries' labour laws and in the regulatory restrictions the
countries place on the types of ®nancial products that can be offered. These effects
are surely complementary to the ones I address here.
21. It is also extremely unlikely that these dif®culties will be overcome by the
issuance of debt and equity in a jurisdiction that offers suf®cient investor
protections. But unless ®rms have assets within these jurisdictions, I do not see
this as a solution.
22. Such developments would be similar to what has happened with the relaxation of
interstate branching regulations in the United States, where banks in one state have
purchased a bank in another state in order to obtain the legal and regulatory
knowledge to do business in that state. Interstate branching has not meant
opening new branches of an existing bank in another region.
Legal Structure, Financial Structure and Monetary Policy Transmission 193
References
Akerlof, G. A., W. T. Dickens and G. L. Perry, (1996) `The Macroeconomics of Low
In¯ation', Brookings Papers on Economic Activity, 1, 1±59.
Angeloni, I., and L. Dedola, (1998) `From the ERM to the Euro: A Soft Transition?', Banca
d'Italia, Working paper, June.
Bernanke, B. S. (1983) `Nonmonetary Effects of the Financial Crisis in the Propagation of
the Great Depression', American Economic Review, 73, 257±76.
ÐÐÐÐ (1993) `Credit in the Macroeconomy', Federal Reserve Bank of New York Quarterly
Review, Spring, 50±70.
Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of
Monetary Transmission', American Economic Review, 82, 901±21.
Bernanke, B. S. and M. Gertler (1989) `Agency Costs, Net Worth and Business
Fluctuations', American Economic Review, 79, 14±31.
ÐÐÐÐ (1990) `Financial Fragility and Economic Performance', Quarterly Journal of
Economics, 105, 87±114.
Calvo, G. A. (1983) `Staggered Price Setting in a Utility-Maximizing Framework', Journal
of Monetary Economics, 12, 383±98.
Cecchetti, S. G. (1995) `Distinguishing Theories of the Monetary Transmission Mechanism',
Federal Reserve Bank of St. Louis Economic Review, 77, 83±97.
ÐÐÐÐ (1996) `Practical Issues in Monetary Policy Targeting', Federal Reserve Bank of
Cleveland Economic Review, 32, 2±15.
ÐÐÐÐ (1998) `Policy Rules and Targets: Framing the Central Banker's Problem', Federal
Reserve Bank of New York Economic Policy Review, 4, 1±14.
Cecchetti, S. G., M. M. McConnell and G. Perez Quiros (1999) `Policy-makers' Revealed
Preferences and the Output±In¯ation Variability Trade-off', Federal Reserve Bank of
New York, February, unpublished.
Cecchetti, S. G. and R. W. Rich (1999) `International Evidence on the Sacri®ce Ratio:
Do We Know How Big It Is?', Federal Reserve Bank of New York, February,
unpublished.
Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited
Participation Models of Money: A Comparison', European Economic Review, 41,
1201±49.
Danthine, J.-P., F. Giavazzi, X. Vives and E.-L. von Thadden (1999) The Future of European
Banking, London, Centre for Economic Policy Research.
de Bondt, G. J. (1997) `Monetary Transmission in Six EU-Countries: An Introduction and
Overview', De Nederlandsche Bank, Research Memorandum 527. Dornbusch, R.,
C. Favero and F. Giavazzi (1998) `Immediate Challenges for the European Central
Bank', Economic Policy, 28, 17±64.
Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy
Transmission Across European Countries', European University Institute Working Paper
98/28, October.
European Central Bank (1999) Possible Effects of EMU on the EU Banking Systems in the
Medium to Long Term, Frankfurt, European Central Bank, February.
Feldstein, M. (1996) `The Costs and Bene®ts of Going from Low In¯ation to Price
Stability', In C. Romer and D. Romer (eds)., Reducing In¯ation, Chicago, University of
Chicago Press, 123±66.
Feldstein, M. (ed.) (1999) The Costs and Bene®ts of Achieving Price Stability, Chicago,
University of Chicago Press for NBER.
Fischer, S. (1977) `Long-term Contracts, Rational Expectations, and the Optimal Money
Supply Rule', Journal of Political Economy, 85, 191±205.
194 Stephen G. Cecchetti
Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from
the G-7 Countries', Centre for Economic Policy Research Working Paper 1219, July.
Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal
of Economics, 95, 43±64.
Grandy, C. (1989) `New Jersey Corporate Chartermongering, 1875±1929', Journal of
Economic History, 49, 677±92.
Groshen, E. and M. Schweitzer (1997) `Identifying In¯ation's Grease and Sand Effects in
the Labor Market', NBER Working Paper 6061, June.
Hubbard, R. G. (1995) `Is There a `Credit Channel' of Monetary Policy?', Federal Reserve
Bank of St Louis Economic Review, 77, May-June, 63±77.
Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G.
Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER.
ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for
the European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives,
September±October, 2±18.
Kieler, M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission? A
Case Not Closed', European Commission Economic Papers, 132, November.
King, R. G., C. I. Plosser, J. H. Stock and M. W. Watson (1991) `Stochastic Trends and
Economic Fluctuations', American Economic Review, 81, 819±40.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants
of External Finance', Journal of Finance, 52, 1131±50.
ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55.
McCauley, R. N. and W. R. White (1997) `The Euro and European Financial Markets',
Bank for International Settlements Working Paper 41, May.
Rotemberg, J. J. (1984) `A Monetary Equilibrium Model with Transactions Costs', Journal
of Political Economy, 92, 40±58.
Taylor, J. B. (1980) `Aggregate Dynamics and Staggered Contracts', Journal of Political
Economy, 88, 1±23.
Vlaar, P. J. G. and H. Schuberth (1998) `Monetary Transmission and Controllability of
Money in Europe: A Structural Vector Error Correction Approach', De Nederlandsche
Bank, May, unpublished.
White, W. R. (1998) `The Coming Transformation of Continental Banking?', Bank for
International Settlements Working Paper 54, June.
Discussion
Manfred J. M. Neumann
195
196 Manfred J. M. Neumann
three partial indicators. Cecchetti then addresses the question whether the
evidence on country differences in credit channel strength can be explained
by differences in the countries` legal systems as regards civil law. Finally ± and
this an innovative aspect of the study ± Cecchetti examines whether the
computed summary index of credit channel strength or importance is
correlated with separate evidence, derived from estimates of structural VARs
on the responses of output and in¯ation to monetary policy shocks. He ®nds
that the expected correlation exists.
While Cecchetti's research agenda is a potentially very fruitful one, the
present study is a ®rst, very tentative exploration. My reading of the evidence
presented is that it does not lend support to the prediction that the impact of
monetary policy is signi®cantly shaped by differences in countries' ®nancial
structures.
1 Data evaluation
For
Banks Concentration Indicator comparison:
per million ratio: of the average
people top ®ve banks importance asset share,
of excluding
small banks top ®ve
Germany 43 17 3 0.02
USA 40 17 3 0.01
Austria 126 48 3 0.05
Denmark 22 17 2 0.74
Italy 16 25 2 0.08
Japan 4 30 2 0.13
France 24 40 2 0.04
Spain 8 44 2 0.18
Greece 2 71 2 1.93
Sources: Cecchetti (this volume), Tables 5.2 and 5.5; own computations.
Kashyap and Stein (1997) have done. But this practice throws considerable
doubt on using this as well as similar indicators as inputs in subsequent
empirical work.
The more types of characteristics are considered, the more relevant becomes
the danger of biased aggregation. Table D5.2 shows the two characteristics
used by Cecchetti to derive an indicator of the availability of alternative
®nance. Countries with the least (most) developed external capital markets
receive an indicator value of 3 (1). Compare, for example, Finland with
Belgium. To be sure, on both characteristics the credit channel should be
stronger in Belgium than in Finland, implying that the rank order should be
reversed. Alternatively, compare Greece with Germany. The ratio of market
capitalisation is smaller in Greece but so is the share of bank loans in total
®nance. Thus, it is not clear on what grounds Cecchetti assumes that the credit
channel is stronger in Greece than in Germany.
Given these inconsistencies in the data transformation, the information
content of Cecchetti's indicators is in doubt. I realise that the researchers of the
credit channel are aware of the crude nature of this work, and I admit that I do
not know how to overcome the methodological problem of aggregating
198 Manfred J. M. Neumann
Austria 15 65 3
Ireland 18 80 3
Greece 20 48 3
Italy 21 50 3
Portugal 23 62 3
Germany 29 55 2
France 38 49 2
Spain 42 58 2
Finland 50 39 2
Japan 67 59 2
Netherlands 96 53 2
Denmark 41 25 1
Belgium 45 49 1
Sweden 99 32 1
United Kingdom 150 37 1
United States 111 21 1
policy on output (though not on in¯ation) and the prediction of relative credit
channel strength derived from his summary indicator of ®nancial structures.
As an additional underpinning Cecchetti also provides a supporting two-stage
least squares regression that employs legal information as instrumental
variables. If valid, this is a nice result. Unfortunately, there is reason to
conjecture that the result is an artefact.
Cecchetti himself notes that the VAR estimates are not very precise,
meaning that the con®dence bands for most countries are very large (see
Ehrman, 1998). Moreover, the estimates suggest for six out of eleven countries
that the maximum impact of monetary policy on in¯ation precedes the
impact on output, in most cases by several quarters. To be sure, this timing is
dif®cult to rationalise and hence throws doubt on the reliability of the
estimates. But let us assume that the VAR estimates for output re¯ect the true
impact of monetary policy. On this assumption Cecchetti's regression appears
to support the conjecture that monetary policy impacts on output are stronger
the stronger is the credit channel. To check for this, Figure D5.1 plots the
output responses against the summary indicator of the credit channel. The
®gure reveals that France and Germany, and possibly the United States, are
0.0
USA
DK P
–0.5 E
Output response
UK S
I
B
IRE
–1.0
G
F
–1.5
0.5 1.0 1.5 2.0 2.5 3.0
Figure D5.1 Maximum impact of an interest rate shock and strength of the credit
channel
200 Manfred J. M. Neumann
outliers. For the remaining eight countries, who belong to different legal
traditions, we observe about the same output response to monetary policy.
In sum, we have to conclude that the evidence collected by Cecchetti does
not support the hypothesis that the effectiveness of monetary policy is
affected by a country's legal tradition or ®nancial structure. However, I hasten
to add that this is not a ®nal verdict on the role of the credit channel as the
negative result might be shaped by the shaky nature of the data employed.
References
Cecchetti, S. G. (2000) `Legal Structure, Financial Structure, and the Monetary Policy
Transmission Mechanism', Chapter 5 in this volume.
Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy
Transmission Across European Countries', European University Institute Working Paper
98/28, October.
Kashyap, A. K. and J. C. Stein (1997) `The Role of Banks in Monetary Policy: A Survey
with Implications for the European Monetary Union, Federal Reserve Bank of Chicago
Economic Perspectives, September±October, 2±18.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants
of External Finance', Journal of Finance, 52, 1131±50.
Discussion1
Ignazio Angeloni
This is a very interesting and stimulating chapter, one that offers a novel
perspective on the transmission process of monetary policy. As in many of his
previous contributions, Cecchetti is able to identify a relevant policy issue,
bring together stimulating evidence and draw intuitively appealing inter-
pretations from it. If the prime test of the success of a study is to stimulate
thinking, this is unquestionably a successful study. No matter how much we
may disagree with the interpretations it proposes and with the overall policy
message ± and, as I will argue below, there is ample room for disagreement ± I
believe that future research on monetary transmision issues, related to the
Euro Area in particular, will have to take the ideas that Cecchetti presents here
carefully into account.
In my comments I will ®rst explain what, in my view, is the main
contribution of this chapter and outline the essence of the argument; I will
then focus on what I regard as limitations of the analysis and explain why and
how these limitations affect the overall policy conclusions.
The chapter builds an ambitious bridge between two so far separate strands
of literature ± that on the bank credit channel of monetary policy transmission
and that on the legal determinants of corporate ®nance and of ®nancial
structure in general. Popularised in the 1990s by contributions by Kashyap
and Stein (1994, 1997), Hubbard (1995), Gertler and Gilchrist (1993), building
on older ideas,2 the `lending view' predicts that the effects of monetary policy
on aggregate demand, output and in¯ation should be in¯uenced (normally
ampli®ed, as shown by Bernanke and Blinder, 1992) by the autonomous
behaviour of commercial banks. The key condition for this to happen is that
bank loans be imperfectly substitutable for other assets and liabilities, both in
the balance sheet of commercial banks and among the forms of ®nancing
available to ®rms and households. The monetary transmission process should
thus be related to the structural (legal, regulatory, cultural, etc.) features of
each country's banking and ®nancial sector that affect the extent to which
direct bank lending can be substituted by other types of credit. As this chapter
convincingly argues, a natural connection arises here with the so-called `legal
201
202 Ignazio Angeloni
view' of corporate ®nance and ®nancial structures, put forward more recently by
La Porta et al. (1997, 1998) and Rajan and Zingales (1998). According to this
literature, cross-country differences in corporate ®nancial structures are related to
the way in which national legal systems accord protection to different classes on
investors. Speci®cally, the `legal view' predicts that better creditor protection will
facilitate external ®nance and help the development of active capital markets, at
the disadvantage of bank intermediation, and that shareholder protection
should make equity ownership more attractive to investors and thus favour
equity relative to debt as a source of capital for non®nancial ®rms.
Having laid down this conceptual premise, the chapter then applies it to the
present situation of the Euro Area. This application is indeed tempting because
the eleven European countries that have adopted a common currency still
display considerable differences in ®nancial structures, banking regulatory
and supervisory frameworks, let alone the legal systems in a broader sense.
Consequently, the chapter's argument goes, the transmission of the single
monetary policy to the Euro Area economy cannot be but diverse across
countries, and will continue to be so unless and until radical steps are taken to
create a common legal framework in the area.
Unfortunately ± or perhaps one should say fortunately for the European Central
Bank (ECB) and the success of its monetary policy ± after closer inspection such
strong policy implication appears less convincing than the basic premise on
which it is built. This is not only because, as one concludes from Table 5.7
(p. 185), legal systems in the Euro Area are not so diverse after all ± the only
country with an Anglo±Saxon legal system is Ireland, and those belonging to the
German or French systems, whose implications for monetary policy effective-
ness are broadly similar, cover nine out of eleven countries or 97 per cent of the
area's GDP ± but for other and more substantive reasons as well. I will organise
my comments on them under three headings: (1) the role of other channels of
monetary transmission (speci®cally, the exchange and interest rate channels);
(2) the changing nature of the bank lending channel in the Euro Area; (3) the
robustness (or lack thereof) of some of the chapter's empirical results.
changes in the central bank controlled interest rates. This stage of the
transmission chain, which crucially depends on monetary policy credibility
and on the ®scal±monetary policy mix, is obviously identical across countries
in a monetary union, while it could differ sharply when each participating
central bank was acting in isolation within its own institutional and national
macroeconomic policy framework. The second step, the impact of exchange
rate changes on the economy, depends on the economy's degree of openness
to international trade. Table D5.3 shows the percentage ratios exports over
GDP in the individual countries of the Euro Area and in the area as a whole,
where in the latter case only extra-area trade is considered. Not only is the
average degree of openness of the area as a whole lower than that of each of the
participating countries, but its dispersion across countries is also signi®cantly
reduced when attention is focused on extra-area trade only. The impact on
individual countries of any given change of the exchange rate of the Euro is
less differentiated than those of each country's exchange rate changes before
monetary union. This argument can be extended to the role that yield curve
changes play in the transmission mechanism. Again, two steps can be
distinguished. First, monetary policy shocks affect the level and shape of the
yield curve, through a mix of liquidity and expectations effects. It has been
shown (Angeloni and Rovelli, 1998) that the impact on the yield curve of
monetary policy changes varies across monetary and ®scal regimes, and that it
is related to monetary policy credibility. Such differences, again, are
eliminated in a monetary union, where a unique yield curve prevails across
the whole area with the only exception of small cross-country spreads that
depend on bond market liquidity and on different credit ratings of sovereign
debtors. The second step is given by the reaction of the private sector's
spending behaviour to yield curve changes. Here the judgement of the
possible effects of the adoption of the single currency is, as in the case of the
exchange rate, more complex. Interest rate effects depend partly on the shape
of private agents' intertemporal preferences, for which no change should
a priori be expected. In addition, life cycle theory suggests also that these
effects should depend also on the size and composition of the household
sector's ®nancial wealth. The income effects of a given interest rate change will
be larger, ceteris paribus, the shorter the ®nancial duration of the household net
®nancial wealth and, for a given duration, the higher the size of wealth. Euro
Area countries have historically been characterised by widely divergent
degrees of liquidity of ®nancial assets, due to different public debt manage-
ment policies.3 As a result of EMU, and in the context of lower and more stable
in¯ation, a convergence process is taking place across Euro Area countries in
the maturity structure of public debts; moreover a convergence of debt ratios
will result from the application of the Stability and Growth Pact. As this
process unfolds, income and substitution effects on consumers following
interest rate changes are likely to follow a more homogeneous pattern within
the Euro Area.
204 Ignazio Angeloni
Note: GDP weights are obtained from national GDP converted into 1997 ECU exchange rates.
According to Kashyap and Stein (1994), three conditions must be present for
such channel to operate. First, monetary policy shocks must affect the overall
size of the banks' balance sheets ± i.e. money demand must be interest elastic;
second, loan supply must respond to money demand changes; third, loans
must be `special' ± that is, not easily substitutable by different forms of
®nancing for the private sector. On all these aspects, signi®cant changes are in
process in the Euro Area.
In recent years, ®nancial innovation has been affecting the nature and
properties of money demand in several countries. New ®nancial instruments,
close substitutes to money, have enabled banks to lower the burden of reserve
requirements and to increase the ¯exibility and diversi®cation of their balance
sheet's liability side. Often, new bank debt instruments have been introduced,
or existing ones have been modi®ed, in response to tax or regulatory changes.
Discussion 205
should be relatively more sensitive to interest rate changes. At least two factors
justify this presumption: the higher relevance of consumer credit and housing
mortgages, that tend to strengthen liquidity and income effects following
short interest rate adjustments, and the higher incidence of shares in private
portfolios, that give rise to stronger wealth effects. The empirical literature is in
agreement with this presumption: for example, the detailed cross-country
study by the BIS (1995) provides convincing empirical backing for it, based on
structural econometric models. Again, this puzzle may partly be explained by
the fact that this chapter concentrates solely on bank lending effects,
disregarding the interest and exchange rate channels.
Second, I found somewhat confusing that the signs of the predicted
in¯uence of the legal family on the effectiveness of monetary policy differ in
sign, in Table 5.8, according to whether the impact on output or in¯ation is
considered. Since all indicators analysed in the chapter refer to effectiveness of
monetary policy in affecting aggregate demand, one would expect the effects
on prices and quantities to have the same sign, and to vary in relative size
according to the nature of the in¯ation±output tradeoff.
4 Conclusions
As I stated at the outset, this is an interesting chapter, fun to read, one that
stimulates thinking. Its central statement, that the legal system may be an
important factor shaping the transmission of monetary policy, is convincing
and justi®es further research.
However, its empirical results and policy conclusions regarding the Euro
Area should be looked at with considerable caution. Relevant areas of the
overall monetary transmission process are not given explicit consideration, in
order to focus speci®cally on bank lending behaviour and on the nature of the
borrower ± lender relationships. Though this simpli®cation may be exposi-
tionally convenient, it should be borne in mind that it could change not only
the size, but also the sign, of the results. Some puzzling empirical estimates
mentioned above may derive precisely from this somewhat unbalanced view
of the transmission process.
The evolution in process in the European banking markets does, in my view,
deserve more attention than it receives here. Restructuring, consolidation and
competition are likely to change considerably the European banking land-
scape and the way bank behaviour affects the transmission process, both in
the aggregate and across different countries, sectors and classes of borrowers.
The `credit channel' in the Euro Area may well have to be studied anew,
building on existing knowledge but taking into account the relevance of the
regime change we are living through. Indeed, for the Eurosystem this may be
one of the most challenging and policy-relevant research issues in the years to
come.
Discussion 207
Notes
1. The views expressed here are the author's, and do not necessarily re¯ect those of the
European Central Bank. I am grateful to Benoit Mojon for useful discussions.
2. The modern literature on the `lending view' has elements in common with the old
one on the `availability doctrine' (see Roosa, 1966). The main difference is that the
new view does not necessarily rely on credit rationing: loan market effects can take
place in equilibrium, through changes in the spread between the loan rate and the
money market rate.
3. I am assuming here that households do not fully internalise future tax liabilities, ±
or, equivalently, that the public debt is at least partly perceived as net wealth by
consumers.
4. According to econometric estimates (Coenen and Vega, 1999) the area-wide M3
aggregate is not controllable in the long run via interest rate changes only.
Controllability is restored if one also considers the effects on output and prices.
5. Since the mid-1990s, a sharp convergence was indeed observed in lending rates
across EMU countries. Such convergence was, however, related not only ± or not
much ± to cross-border competition in lending, but to a large extent to the general
convergence of in¯ation and interest rates.
References
Angeloni, I. and Rovelli, R. (1998) Monetary Policy and Interest Rates, London, Macmillan.
Bank for International Settlements (BIS) (1995) Financial Structure and the Monetary Policy
Transmission Mechanism, Basle, BIS, March.
Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of
Monetary Transmission' American Economic Review, 82, 901±21.
Coenen, G. and J. L. Vega (1999) `The Demand for M3 in the Euro-Area', European
Central Bank, Working Paper Series 6, September.
Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal
of Economics, 95, 43±64.
Hubbard, R. G. (1995) `Is There a ``Credit Channel'' of Monetary Policy?', Federal Reserve
Bank of St Louis Economic Review, 77, May±June, 63±77.
Kashyap A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G.
Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER.
ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for the
European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives,
September±October, 2±18.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (1997) `Legal
Determinants of External Finance', Journal of Finance, 52, 1131±50.
ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55.
Rajan, R. G. and L. Zingales (1998) `Which Capitalism? Lessons from the East Asian
Crisis', Bank of America, Journal of Applied Corporate Finance, 11, 40±8.
Roosa, R. V. (1966) Interest Rates and the Central Bank; reprinted in Richard S. Thorn,
(ed.), Monetary Theory and Policy, New York, Random House, 559±83.
6
1 Introduction
There are three main issues which must be discussed if one wishes to answer
the question which this chapter addresses:
(1) Which features of a ®nancial system are important for monetary policy
or, in other words, how is monetary policy conducted, and how does it
affect the real economy, and how and to what extent does this depend on
the speci®c features of an economy?
(2) How different are the ®nancial sectors ± or more generally the ®nancial
systems ± in Europe?
(3) If signi®cant differences existed between countries, would this have
consequences for how monetary policy should, and can, be conducted in
a common currency framework?
208
Differences Between Financial Systems in European Countries 209
monetary aggregate) in the economy and for the banking sector. To the extent
that banks re®nance themselves, or hold voluntary or compulsory reserves, at
the central bank, they have fewer funds to offer to their potential borrowers
and/or lending becomes more costly to them. The resulting change in the
situation of the banking sector has, in turn, consequences for the real
economy to the extent that the ®rms, households and government bodies that
want to borrow from the banks have less credit available, or have to pay more
for the funds borrowed from the banking sector. The restrictive monetary
policy impulse thus works itself through the system and determines aggregate
demand and ultimately output and income, and possibly the price level as
well. Some details of this process have to be left for further discussion, but
evidently I assume that monetary policy is effective, at least in the short run.
The second idea is taken from a paper by Dornbusch, Favero and Giavazzi
(1998). Differences between the national ®nancial systems might matter
because different national economies might react differently to monetary
policy impulses of a given kind and size. If this is so, it might be so because of
differences in the ways in which central bank action in¯uences the respective
®nancial sectors and in particular the banking sectors, or because of differences
in the ways in which a change in the situation of the ®nancial sectors affects
the real economy. If there are two countries, A and B, in a monetary union,
with A being more sensitive to monetary policy impulses than B, then a given
monetary policy measure might be too strong for country A and too weak for
country B even if it is just right for the intended effect on the (weighted)
average of the two economies in terms of, say, their in¯ation rates.
If I may claim any comparative advantage at all for discussing the
consequences of national differences for EMU, it has to do with the second
question ± that is, the differences between the ®nancial systems in Europe.2
However, here I must also make two reservations:
From the way I have formulated the title of my chapter and the questions
listed above, it can be seen that I will not con®ne myself to looking at
differences between ®nancial sectors, but wish to ± and will later show that I
need to ± discuss differences between national ®nancial systems. This is not
simply a terminological difference, but a difference in substance. I use the term
`®nancial sector' in a narrow way: it denotes those specialised institutions such
as banks, pension funds, securities markets, etc. which provide ®nancial
services to the non-®nancial sectors of the economy, and to the ways in which
these institutions are shaped and managed and how they operate and are
regulated. The term `®nancial system' encompasses not only the ®nancial
sector, but also the real sectors to the extent that they demand the ®nancial
services of the ®nancial sector and also to the extent to which they forgo using
the ®nancial sector, as well as the interaction between the demand for and the
supply of the services of the ®nancial sector. Thus, for instance, the extent to
which internal ®nancing of investment takes place, the extent to which saving
takes the form of real investment, the extent to which banking services are
210 Reinhard H. Schmidt
appropriate to the demand for them, etc. are features of a given ®nancial
system.3
The second reservation refers to the countries which I will take into
consideration. Anticipating a likely future course of events, I will assume that
the United Kingdom is already part of the European Monetary Union (EMU).
Moreover, my empirical references will be mainly restricted to the three largest
®nancial systems in the `enlarged EMU' ± namely those of France, Germany
and the United Kingdom. Even if the Italian economy were larger than that of
Great Britain, I would wish to include that of the United Kingdom because, as I
will describe below, its ®nancial system differs markedly from those of the
continental European economies, and from that of Germany in particular. An
understanding of these differences helps to underscore the heterogeneity that
makes for national diversity, and thus for possible problems of a common
monetary policy, in the EMU.
The chapter is structured as follows: In Section 2, I shall discuss the
transmission mechanisms. Given my background, it appears almost self-
evident to me that the so-called `credit channel' is important,4 though not so
much as an alternative to the `conventional' channels of the transmission of
monetary policy, but rather as a complement to them. The discussion of the
various channels serves the purpose of providing the criterion for selecting
those aspects of national ®nancial systems which I will then go on to
characterise in some detail. In Section 3, I will argue that the ®nancial systems
± as well as the ®nancial sectors ± of the three countries are vastly different, and
point out the main differences. In Section 4, I will bring the two lines of
reasoning together and argue that, in spite of what I perceive as important
differences, I do not think that these differences matter very much for the
design and conduct of a common monetary policy. The main reason for this is
that what I consider to be the characteristic features of the different ®nancial
systems ± though not of the different ®nancial sectors ± might have two effects
on the functioning of monetary policy which tend to offset each other.
The interest rate channel (a) and to a certain extent also the channel of
relative prices (b) are standard elements of what may be regarded as the
traditional view of the transmission mechanism. The exchange rate channel
(c) is not relevant in the context of the present chapter, as the ± enlarged ±
EMU area is a very large economy for which exchange rates are not a terribly
important factor because external trade accounts for only a relatively small
share of total GDP. The credit channel (d) is the `newcomer' in the market of
ideas and by now the main `competitor' to the incumbents (a) and (b).
Financial System
(1) (2)
Financial Financial
sector patterns
(Role of (Saving and
institutions and financing)
market)
(3) (4)
Corporate Corporate
governance strategy
(Roles of various (Ability regarding
stakeholders) fundamental
change)
one can assume that it has positive consequences if this is the case, and
negative consequences if the elements do not ®t. One can speak of `important'
differences between ®nancial systems, if not only the main elements of which
they are composed, but also the way in which they are related, are different.
The interesting thing about the sub-systems and their interrelationship,
and thus the interesting thing about the entire ®nancial system, is that they
are composed of complementary elements. Elements of a system are called
`complementary' (to each other) if they mutually increase the `bene®t' they
yield in terms of whatever the objective function or the standard for
evaluating the system may be, and mutually reduce their disadvantages or
`costs'.15 A system is called consistent if its complementary elements indeed
take on the values which make the system attain a local optimum. Systems
of complementary elements typically have more than one optimum, and
the local optima are clearly distinct con®gurations of the values of the
elements.
Financial systems are systems in this speci®c sense. The complementarity of
their elements and the economic bene®ts which a consistent ®nancial system
can be assumed to produce are the factors which account for the tendency of
countries' ®nancial sectors and non-®nancial sectors to co-vary in a systematic
way. To illustrate the concepts of complementarity and consistency in the case
of ®nancial systems and also to lay the foundation for the next argument, I
would now like to take a look at common two-way classi®cations from the
literature for the four sub-systems shown in Figure 6.1.
however, I hope that by drawing attention to a few key aspects I will be able to
make my point. It is equally clear that one cannot expect to ®nd a perfect
correspondence between a given type of ®nancial system and a given real
®nancial system ± i.e. that of a speci®c country.
Nevertheless, the classi®cation of ®nancial systems suggests differences in
reality.22 The example which immediately come to mind are the German and
the British ®nancial systems ± and this is the main reason why I anticipate a
likely future course of events and assume that the United Kingdom is already
part of `Euroland'.
The German ®nancial system is of the ®rst type. The German banking sector
is big and some banks are important and powerful, whereas capital markets
and capital market-oriented institutions such as pension funds are `under-
developed', as can be measured by indicators such as the total assets of banks
and stock market capitalisation as percentages of GDP, respectively.23
Corporate governance in Germany functions mainly through internal
mechanisms and involves `insiders' to the corporations ± or, in other words,
people and institutions which have long-term interests in safeguarding their
speci®c relationships with a corporation and that are typically better informed
about its prospects and problems than anonymous market participants could
be.24 There does not seem to be an active market for takeovers. The role of
banks, of other corporations and of employees having codetermination rights
in the governance of corporations epitomise this system, and despite the
widely publicised declarations to the contrary, the maximisation of share-
holder value is not the dominant objective of most large German ®rms. Even
though this may be more speculative, one can also add the observation that, at
least on an economy-wide level, German ®rms are more woven into nets of
implicit contracts, that labour turnover is lower and that the corporations are
less able and willing to undertake abrupt adjustments to changing circum-
stances than British corporations. Thus it is fair to say that at least as regards
the ®rst, the third and, to a certain degree, the fourth element (or sub-system),
the German ®nancial system is consistent.
Without doing too much violence to a reality which is of course much more
complex, one can characterise the British ®nancial system as one in which
core elements ± or sub-systems ± (1) and (3), and possibly also (4),25 take on
values which are precisely the opposite of those determined for Germany. In
the United Kingdom, the relative importance of banks in the ®nancial sector is
not as great; capital markets and institutions ± notably pension funds, which
are capital market-oriented and in most cases independent of the banks ± play
an important role; there is no formal and no de facto codetermination;
and ®nancial intermediaries are scarcely involved in the governance of
corporations.26 The corporate objective is unrestricted maximisation of pro®ts
or shareholder value, and given the nature of the UK ®nancial system there is
also no reason why this should be otherwise. The basic mechanism of
corporate governance is the takeover market, and the entire corporate
220 Reinhard H. Schmidt
BE Higher Lower BE
Germany
FS FP Japan
– Households hold
– Banks play primarily bank deposits
– Bank loans as main
Lower
Lower
dominant role
financial source of firms
Limited
coherence
CG CS
System-wide role of market mechanisms
– Control – Firms-specific
– Relationships investments
– Inside influence – Continuous
– Insider control adjustments
USA
FS FP
UK
– Organised – HHs hold claims
capital markets on funds and securities
play dominant – Market-based
role firm financing
Higher
Higher
Limited
coherence
CG CS
– Liquidity – Marketable
– Arm's length investment
– Outside option – Fundamental
– Outsider strategic shifts
control
BE Higher Lower BE
which do exist, are inconsistent with what one would expect on the basis of
theory and of the empirical features of sub-systems (1) and (3) in the various
countries. For instance, Mayer (1988, 1990) found in his early studies that
bank ®nancing is more important in the United Kingdom than in Germany
and that equity ®nancing is even negative in the United Kingdom. Corbett
and Jenkinson (1996, 1997) supported these results in more recent and much
more extensive studies. These ®ndings do indeed pose a puzzle. If they were
true, they would break the logical chain connecting the four sub-systems and
contradict the proposition that the German and the British ®nancial systems
are consistent. Moreover, they would invalidate the whole concept of ®nancial
systems as consistent sets of complementary elements.
From the perspective adopted in the present chapter, it is fortunate that the
methodology developed by Mayer and used by him and others has a ¯aw.
Mayer and his followers look at net ¯ows of funds between economic sectors
whereas they should have looked at gross ¯ows, because in the process of
aggregating over real and ®nancial investments and over time, netting
eliminates the relevant differences which might exist, and thus the role of
external debt ®nancing disappears almost completely. In his dissertation,
Andreas Hackethal (1999) has identi®ed this ¯aw, suggested a way to avoid it
and recalculated the ®nancing structures of ®rms in Germany, Japan and the
United States. As the ®rst columns per country (with the heading `gross') in
Table 6.1 show, his correction yields patterns of long-term ®nancing which are
consistent with expectations based on the structure of the entire ®nancial
systems. The second columns (`net') and the ®rst line concerning internal
Table 6.1 Financing patterns in various countries, 1970±96, per cent of physical
investment (gross: 1970±96; net: 1970±94)
Note:
Net ®gures and those for `Internal funds' are taken from Corbett and Jenkinson (1997). They add up
to 100 per cent. These authors include both `NBFI ®nance' and `Commercial paper' in `Bank
®nance'(except for the United States where `Commercial paper' is included in `Others').
Gross ®gures, which by construction do not add up to 100 per cent, are taken from Hackethal (1999).
funds, for which the net and gross ®gures are identical, show the correspond-
ing net ®gures derived by Corbett and Jenkinson (1997) for comparison.33
The data needed to recalculate the ®gures for French and British ®rms, and
thus to provide a more accurate picture of their ®nancing patterns, are not yet
available, but it seems plausible that, given the similarity of the ®nancial
systems of the United States and the United Kingdom, those for the United
Kingdom will not differ too much from those for the United States. If this is
indeed the case, the claim that the British and the German ®nancial systems
are consistent and differ in a fundamental way can be upheld.
In another empirical study we have found further interesting evidence to
support the assumption of fundamental differences between the British and
the German ®nancing systems and the proposition that dramatic changes are
under way in France.34 In this paper, we have calculated various intersectoral
intermediation and securitisation ratios for the three countries over a span of
®fteen years. In contrast to the `conventional wisdom' that there is a general
tendency towards disintermediation and securitisation and that, overall, the
®nancial systems in Europe are becoming more similar, our study shows that
the levels of disintermediation and securitisation differ substantially between
countries and that, except for France, they are surprisingly stable over time. As
Figure 6.3 on p. 224 shows, the `liability±intermediation ratios of non-
®nancial companies', which measures the fraction of external ®nancing which
comes from intermediaries, (Figure 6.3a) and the `liability-intermediation
ratios of non-®nancial companies to banks', which measure the share of bank
®nancing in the total external ®nancing of ®rms (Figure 6.3b), and the ratios of
securitisation of corporate ®nancing (Figure 6.3c) differ greatly between
Germany and Great Britain, and are almost completely stable in these two
countries. For France, these ratios exhibit not only a great instability, but also a
tendency to move away from the German to the British model.
This may suf®ce to demonstrate that the general characterisations of the three
®nancial systems are empirically valid, that the overall ®nancial systems of
Germany and the United Kingdom are consistent con®gurations of the sub-
systems as complementary elements and that there are indeed considerable and,
at least in some cases, persistent differences between the ®nancial systems of the
major European countries. Having established this, I can now analyse what
consequences this might have for a common monetary policy within EMU.
a) Liability–intermediation ratios
a) Liability–intermediation of non-financial
ratios companies
of non-financial (NFCs)
Companies (NFCs) b) Liability–intermediation ratios of
of NFCs
banks to banks
Financial liabilitiesofNFCs
Financial Liabilities NFCstotoFS/
FS/ T
Total Financial
financial Liabilities
liabilities of
of NFCs
NFCs Financial liabilities of
Financial Liabilities NFCs
NFCstotobanks/
banks/ TTotal financialLiabilities
otal financial liabilitiesofofNFCs
NFCs
85 80
80 Germany 70
70 Germany
75 60
60
70
65 50
50
France
60 40
40
United Kingdom
55 30
30
50 20
France 20
45
United Kingdom
40
10
35 0
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
c)c)Liability–securitisation ratios
Liability–Securitisation of of
ratios non-financial companies
Non-financial companies d)
d) Liability
Liability [Asset]–intermediation
[Asset]–intermediation ratios
ratios of
of banks
banks
Securitised
itisation liabilities FinancialLiabilities
Financial liabilities[assets]
[assets]of of banks
banks to [from]
to [from] NBFIs/
NBFIs
Secur Liabilities of of NFCs
NFCs / Total
/ Total financial
financial liabilities
Liabilities of NFCs
of NFCs
Totalfinancial
Total financialliabilities
liabilities [assets]
[assets] of NFCs
of banks
80
United Kingdom 20 Liability–intermediation ratios
70
France
60 15
France
50 Germany
40 10
30 d United Kingdom
Germany 5
20
10
A-IRs
0 0
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
Figure 6.3 Intermediation and securitisation ratios (in percent), 1981±96
differ; and (2) the FSecs may be, and indeed are, dissimilar; and (3) the ways in
which the ®nancial sector changes credit supply to ®rms and households, and
in which the real sectors change their demand for ®nancial services, in
reactions to the changing conditions in the ®nancial sector (B) seem to be
different; and (4) important aspects of RE ± notably company ®nancing,
governance and strategies and structures ± are likely to differ. Or to put it more
succinctly: ®nancial and real sectors and the sensitivities of the ®nancial sector
to central bank policies, and of the real sectors to changing conditions in the
®nancial sector may differ between countries.
Comparing Germany and the United Kingdom, one can ask: in which
system does a central bank impulse of a given kind and size have a stronger effect
on the ®nancial sector (A)? And in which of the two countries does a given
amount of in¯uence of the central bank on the ®nancial sector have a stronger
effect on the aggregate demand of the real sectors (B) ± and ultimately on
income, employment and prices?
The answers depend on how one sees these relationships on the structures of
the ®nancial systems and in particular on the roles of the banking sector in the
respective ®nancial system. I use the concepts underlying the notion of a
credit channel. Here the immediate effect of central bank action is assumed to
be mainly an effect on the banks. Thus, the strength of the relation A between
the central bank and the ®nancial sector is likely to be stronger in a country in
which the relative importance of banks in the ®nancial sector is greater. This
should be the case in Germany because of the following main reason. The
German ®nancial sector consists mainly of universal banks which supply
nearly all kinds of ®nancial services under one roof.35 In contrast to the British
®nancial sector banks do not have any serious NBFI-competitors in
Germany.36 Thus the central bank can determine more directly the
re®nancing conditions of a much larger part of the ®nancial sector. I therefore
hypothesise that the ®nancial sector of Germany reacts more strongly to
monetary policy impulses than the British ®nancial sector.
As a second step, one has to ask how changing conditions in the entire
®nancial system ± and, in particular, in the banking sector ± affect decisions of
®rms and households. To provide a concrete example, let us examine the case
of a tightening of monetary policy.37 Would funding opportunities and
funding costs for the real sectors be restricted or raised more in Germany or in
the United Kingdom? I presume that the restrictive effect of a given in¯uence
of the central bank on the ®nancial sector ± and, in particular, on the situation
of the banks ± would be less in Germany than in the United Kingdom, and this
for three reasons.
The ®rst reason is that the relationship between banks on the one side and
®rms and households on the other is closer in Germany than in the United
Kingdom. Owing to their closer multifaceted and more long-term-oriented
relationships with their customers, German banks might be more hesitant to
tighten credit terms for their clients in order not to burden them too much and
226 Reinhard H. Schmidt
the central bank or not. At the level of large companies, a strict and exclusive
short-term pro®t orientation41 is also the exception and not the rule in
Germany. The more moderate pro®t orientation of many large corporations
tends to make them `ignore' changing ®nancial conditions so as to promote
their long-term strategies, and they probably are more prone to do this than
their counterparts in the United Kingdom. Here I would like to reiterate that
this behavioural pattern on the part of German banks and the large ®rms is not
necessarily a `problem' or a `¯aw'; rather, it is a feature that is consistent with
the general logic of the German ®nancial system.42
In the United Kingdom, both banks and corporations are more forced to
strive for pro®t and in so doing react to changing price signals. Both in their
relationships with clients and in the way they value outstanding securities,
privately owned and unambiguously pro®t-oriented banks and other ®nancial
intermediaries react more strongly to central bank activity; and private ®rms
which have to be extremely conscious of their stock price performance are
more strongly motivated ± indeed, forced ± to react more to price signals, as it
is to be expected in a more market-oriented system.
Thus one can conclude that the two systems also have different reaction
functions with respect to the relationship between the ®nancial sector and the
non-®nancial sectors (B). The British system reacts more, and the German
system reacts less, to changes in the situation of the respective ®nancial
sectors. Figure 6.4 summarises the two parts of the argument presented so far.
As can be seen, the German system combines a stronger in¯uence of the
central bank on the situation of the ®nancial sector (A, with four stars in the
second line of Figure 6.4) and a weaker in¯uence of the ®nancial sector on the
real economy (B, with two stars). The situation in the United Kingdom is
the reverse. The dominance of the central bank over the ®nancial system is
weaker (A, with two stars in the third line) and that of the ®nancial sector over
the real sector is stronger (B, with four stars).
It would, of course, be naive to quantify the strength of these effects solely on
the basis of the arguments which I have provided. Accordingly, all that I would
like to do here is to present the idea that the relative strengths of the two
relationships A and B are inversely related. That this should be so, is not by
coincidence. Instead, it follows from the systemic features of the ± broadly
de®ned ± ®nancial systems, which are consistent in each of these two countries.
If one carried the exercise of `counting stars' to an extreme, one would ®nd
that the net effect of monetary policy may be the same in both countries under
consideration here. But I do not mean to imply that the two differences
between the partial channels A and B in the two countries cancel
out completely. I chose to use stars, and not numbers, because the latter might
have suggested a greater degree of precision, perhaps leading to
misunderstandings.
However, it does seem safe to assume that the differences between the
stylised ®nancial systems have offsetting effects with respect to A and B, and in
228 Reinhard H. Schmidt
A B
CB FSec RE
Credit Germany
channel
United Kingdom
Figure 6.4 Monetary policy effectiveness in Germany and the United Kingdom
itself this may be an important factor for assessing the potential for a common
monetary policy in an enlarged `Euroland'. The two ®nancial systems may
differ in a fundamental way, and yet the net effectiveness of monetary policy
may not be all that different. If this is news, it is good news for the proponents
of a monetary union in Europe and its future enlargement.
The argument presented here in a very non-technical manner implicitly
draws on the credit channel literature. It could be rendered more precise and
more technical by making this basis explicit and spelling out in some detail
how the differences between the ®nancial systems determine the strength of
the relationships A and B in the two countries. Not unexpectedly, this exercise
would demonstrate that the effects of institutional features on the transmis-
sion mechanism are complex, dif®cult to aggregate and in some cases
ambiguous. Nevertheless, the basic thrust of the argument would not change;
the overall effects of the working of the bank lending and the balance sheet
channels are in line with my intuitive argument.43 But even a more technical
approach would not solve one important aggregation problem: How can one
`add' the different effects of the links A and B given that the relevance of the
link B depends on how strong the link A is? It seems that this question requires
much more work and in-depth econometric studies. I would be delighted if the
brief non-technical discussion I have presented here were to be the inspiration
for such a research project.
Having admitted that my attribution of stars to the partial channels A and B
in the two stylised ®nancial systems owes a lot to the credit channel literature,
I should brie¯y add a conjecture concerning whether the result is stable with
Differences Between Financial Systems in European Countries 229
respect to the theoretical basis. The general thrust of the credit channel
argument is that the possible effects of monetary shocks, including those
engineered by a central bank, are probably greater than those which the
literature based on the interest rate channel takes into account. If one retains
the notion of the ®nancial system as a system of complementary elements,
and the premise that Germany and the United Kingdom have consistent but
differing ®nancial systems, and combines them with the assumptions of the
interest rate channel, then the effects for both partial channels would be
weaker: the effect of the central bank on the banking system (A), which could
be approximated by the multiplier, would be less pronounced in both
countries and the difference between the effect in Germany and the effect in
the United Kingdom would still be about the same. The effect of a change in
the situation of the banks and other ®nancial institutions on the real economy
(B) would probably also be weaker in general ± and at least some of the
arguments which would suggest that in the United Kingdom this effect is
stronger would also still apply. Thus, the main result ± namely, that the two
parts, A and B, of the transmission mechanisms tend to exhibit offsetting
differences between the two countries ± would also be valid if one disregarded
the speci®c aspects of the credit channel. This is shown by the stars bracketed
on the lower lines in Figure 6.4.
change in general and what this implies for the common monetary policy
in `Euroland'. Owing to space constraints, I will have to restrict myself to a
few brief remarks.
The `fact' that ®nancial systems have the property that every element of a
®nancial system is functionally related to many others and that a consistent
®nancial system constitutes a local optimum is relevant here. It implies that
partial reforms are not likely to be sustainable, and this in turn implies that
®nancial systems are not likely to change and to converge gradually.
But if there is no gradual convergence, what else could happen? If, for
instance, political forces or dynamic entrepreneurs in the ®nancial service
industry succeed in introducing elements of one system ± for instance, the
active takeover market of the British system ± into the other ± for example, that
of Germany ± the immediate result would be an inconsistent `non-system'. If
this happens, there will be pressure to restore consistency. A `restauration' of
the old system is one way of achieving consistency. In this case the
`innovation' is rejected. This is one possible course of events.
The other possibility is that the forces which make changes appear
desirable are strong and the forces of `restauration' weak and that the
formerly consistent system of a given country undergoes not only a series
of partial changes, but will in fact experience a `Gestalt switch' from one
type to the other. The resulting new con®guration of the elements of the
system may be better or worse; one local optimum is replaced by another
one.44
In both cases ± that is, in the case of partial reforms which only lead to a
`restauration' and in the case of a complete transformation of the system's
architecture ± the task of monetary policy makers will be extremely dif®cult
because the transmission mechanism will have become unstable. In such a
situation the European Central Bank would scarcely be able to predict the
overall effects of its policy on the economy of `Euroland' and would
therefore ®nd it dif®cult to determine the precise strength of the monetary
impulses which it should provide. There would be less of a chance of such a
destabilisation and loss of orientation of monetary policy makers occurring
in Europe as a consequence of changes in the national ®nancial systems if
different national currencies and monetary policies had been retained and
not replaced by a uni®ed currency regime; and under the old system of
national monetary policies, such a destabilisation would have less serious
consequences. I want to conclude by expressing my belief that this
increased risk of instability and disorientation constitutes a bigger problem
for a common currency than the need to design and implement a common
monetary policy for different, but essentially stable, ®nancial systems, on
which the existing literature focuses.
Differences Between Financial Systems in European Countries 231
Appendix
Notes:
a
For UK authorised banks.
b
UK building societies.
Sources: Wymeersch (1998); Towers Perrin's 1998 Worldwide Total Remuneration Report.
Differences Between Financial Systems in European Countries 233
(a) Corporations
Dividends/cash ¯ow ratio for listed
companies (1994) (%) l2.7 l6.67 9.46
Corporate loans collateralised by real
estate as percentage of total corporate loan (1993) 36 59 4l
Short-term corporate credit
as percentage of total corporate loan (1993) 22 50 27
Long-term corporate credit
as percentage of total corporate loan (1993) 78 50 73
(b) Households
Mortgage credit interest adjustment (%)
Fixed 20 0 80
Renegotiable 40 30 0
Variable 0 0 20
Reviewable 40 70 0
Typical loan±value ratio 60±80 90±95 70±80
Notes:
Renegotiable: rate not ®xed over entire term, but more than one year
Sources: Maclennan, Muellbauer and Stephen (1998); Borio (1995); La Porta et al. (1999)
Notes
1. The author wishes to thank Falko Fecht, Andreas Hackethal and Adalbert Winkler,
and in particular Marcel Tyrell, for extremely valuable advice and research support.
Financial support of the Deutsche Forschungsgemeinschaft is gratefully acknowl-
edged.
2. There is a growing body of econometric literature on differences regarding the
monetary transmission mechanisms in European countries (see for instance Bank
for International Settlements, 1995; Dornbusch, Favero and Giovazzi, 1998; Britton
and Whitley 1997; Ramaswamy and Sloek, 1997; Giovanetti and Marimon 1998).
To me this literature, which is partly surveyed in Dornbusch, Favero and Giovazzi
(1998), does not provide unambiguous evidence of great differences in the
transmission mechanisms. However, to the extent that these papers discuss
differences between national ®nancial systems at all, they do not go very far in this
respect. But see Kashyap and Stein (1997a), who follow a similar approach to the
one in the present chapter.
3. See Schmidt and Tyrell (1997) for this terminological distinction.
4. See Schmidt (1990).
5. See for instance the articles by Bernanke and Gertler (1995); Meltzer (1995);
Mishkin (1995) and Taylor (1995) in the Journal of Economic Perspectives and
symposium on monetary transmission mechanism, and Goodhart (1989);
Cecchetti (1995) and Illing (1997, pp. 145 ff.) for overviews.
6. Especially to households' decisions concerning the acquisition of homes and of
consumer durables.
7. See Borio (1997) for an exhaustive analysis of different monetary policy procedures
and their recent convergence.
8. See Meltzer (1995).
9. See Bernanke and Gertler (1995) for this argument.
10. For surveys of this transmission channel with empirical results for the United
States, see also Bernanke, Gertler and Gilchrist (1996) and Kashyap and Stein
(1997b).
11. See Bernanke, Gertler and Gilchrist (1996).
12. For an overview see Freixas and Rochet (1997). More recent contributions include
Rajan (1996) and Kashyap, Rajan and Stein (1999). The quotation paraphrases the
title of an in¯uential article by James (1987).
13. So far, and according to my knowledge, the credit channel has been the subject of
only a few empirical investigations in Germany. Sto È û (1996) and Guender and
Moersch (1997) come to a negative conclusion concerning the importance of the
bank lending channel in Germany. Worms (1997) ®nds some positive evidence
with respect to the balance sheet channel. Kuppers (1998a, 1998b) forcefully
criticises the results of Guender, Moersch and Sto È û and ®nds strong support for a
credit channel in his own empirical study. For the United Kingdom Dale and
Haldane (1995) and Ganley and Salmon (1997) show some importance of the
credit channel. More recent research on the credit channel in France includes Goux
(1996), Candelon and Cudeville (1996) and Payelle (1996). Their results are
somewhat ambiguous, but support the assumption that the credit channel is
relevant in France too.
14. The following discussion is based on Hackethal and Schmidt (2000). See also
Milgrom and Roberts (1995); Hackethal and Tyrell (1998) and Aoki (1999).
15. For this de®nition see also Milgrom and Roberts (1995). The mathematics behind
the concept of complementarity are surveyed by Topkis (1998).
Differences Between Financial Systems in European Countries 235
16. See Rybczinski (1984) and Berglof (1990) for this classi®cation with respect to the
®rst and second sub-system.
17. These dichotomies and the way in which they are related, are discussed in
Hackethal and Schmidt (2000).
18. See Franks and Mayer (1994); Schmidt (1997b) and Tirole (1999).
19. See Boot and Macey (1998), Hackethal and Schmidt (2000); Aoki (1999).
20. The link between bank-oriented ®nancing, insider-controlled governance and ®rm-
speci®c human capital is more deeply analysed in Hackethal and Tyrell (1998) and
Berkovitch and Israel (1998). The correspondence to the business systems is
discussed by Milgrom and Roberts (1995); Aoki (1999) and Hackethal and Schmidt
(2000). See also Mayer (1998).
21. See Williamson (1988).
22. See also Goodhart (1993).
23. See the empirical results in European Central Bank (1999) and Davis (1998). Some
indicators are presented in the Appendix.
24. See Schmidt and Tyrell (1997) and Prigge (1998).
25. For the United Kingdom we do not know enough about (4), but see Hackethal and
Schmidt (2000) for ®rst results and some rather speculative conclusions.
26. See Goergen and Renneboog (1998) and Franks, Mayer and Renneboog (1998).
27. See Charkham (1994); Franks and Mayer (1997) and Wymeersch (1998).
28. For a similar characterisation of the British system, see also Prevezer and Ricketts
(1994).
29. See Schmidt (1997a) for some details. The recent book by Plihon (1998, p. 79), is
among the numerous supporting French references which one could quote here.
30. For a similar conclusion see OECD (1995).
31. For instance, this can be concluded from the more active market for corporate
control in recent years (Wymeersch, 1998). On the other hand, ownership
concentration and voting power in French public corporations indicates an insider-
control system (Bloch and Kremp, 1998).
32. See Mayer (1988, 1990) and Corbett and Jenkinson (1996, 1997) for international
comparisons of ®nancing patterns and Edwards and Fischer (1994) for a study of
Germany and Bertero (1994) for France.
33. In the work of Mayer and his followers, one can ®nd another distinction. It is the
distinction between net and gross ®gures, which concerns a different aspect from
the one under discussion here. Their 'gross ®gure' are calculated after the
aggregation which is identi®ed in Hackethal (1999) and Hackethal and Schmidt
(1999) as the cause of the bias.
34. See Schmidt, Hackethal and Tyrell (1999).
35. See for instance the detailed descriptions of the British and German ®nancial
sectors in Saunders and Walter (1996).
36. The predominance of the universal banks in Germany may be due to regulatory
conditions partly resulting from the monetary policy strategy of the Deutsche
Bundesbank, which heavily depended on a stable money demand. See for instance
the Deutsche Bundesbank (1995).
37. This is more than a way of making the discussion more concrete. It might well be
that the arguments which follow, only apply to the special case of a restrictive
monetary policy.
38. This re¯ects that there is some truth to the belief that the old system of
'housebanks' still prevails. This has been vigorously challenged in a well known
book by Edwards and Fischer (1994). But note that the empirical basis of their
236 Reinhard H. Schmidt
References
Allen, F. and D. Gale (1997), `Financial Markets, Intermediaries and Intertemporal
Smoothing', Journal of Political Economy, 105, 23±46.
Aoki, M. (1999), 'An Information Theoretic Approach to Comparative Corporate
Governance', Department of Economics, Stanford University, Working Paper 99±04.
Bank for International Settlements (1995), Financial Structure and the Monetary Policy
Transmission Mechanism, Basle: Bank for International Settlements.
E. (1990), `Capital Structure as a Mechanism of Control: A Comparison of
Berglof,
Financial Systems', in M. Aoki, B. Gustafsson and O. E. Williamson (eds), The Firm as a
Nexus of Treaties, London: Sage, 237±62.
Berkovitch, E. and R. Israel (1998), 'Optimal Bankruptcy Laws Across Different Economic
Systems', University of Michigan, Working Paper.
Differences Between Financial Systems in European Countries 237
Bernanke B. and M. Gertler (1995), `Inside the Black Box: The Credit Channel of
Monetary Policy Transmission', Journal of Economic Perspectives, 9, (4) 27±48.
Bernanke, B., M. Gertler and S. Gilchrist (1996), `The Financial Accelerator and the Flight
to Quality', Review of Economics and Statistics, 78, 1±15.
Bertero, E. (1994), `The Banking System, Financial Markets, and Capital Structure: Some
New Evidence from France', Oxford Review of Economic Policy, 10, (4),68±78.
Bloch, L. and E. Kremp (1998), Ownership and Voting Power, Louvain, European Corporate
Governance Network.
Boot, A. W. A. and J. R. Macey (1998), 'Objectivity, Control and Adaptability in
Corporate Governance', Tinbergen Institute, Discussion Paper, 98±064.
Borio C. E. V. (1995), `The Structure of Credit to the Non-Government Sector and the
Transmission Mechanism of Monetary Policy: A Cross-Country Comparison', in Bank
for International Settlements (ed.), Financial Structure and the Monetary Policy
Transmission Mechanism, Basle: BIS, March, 59±105.
ÐÐÐÐ (1997), `The Implementation of Monetary Policy in Industrial Countries: A
Survey', BIS Economic Paper, 47.
Britton, E. and J. Whitley (1997), `Comparing the Monetary Transmission Mechanism in
France, Germany and the UK: Some Issues and Results', Bank of England Quarterly
Bulletin, May,152±62.
Candelon, B. and E. Cudeville (1996), `Politique monetaire
et canal de credit: une
estimation empirique sur l'economie
francËaise', Revue d'economie politique, 104,
781±807.
Cecchetti, S. G. (1995), `Distinguishing Theories of the Monetary Transmission
Mechanism', Federal Reserve Bank of St. Louis Review, May±June, 83±97.
Charkham, J. P. (1994), Keeping Good Company: A Study of Corporate Governance in Five
Countries, Oxford, Oxford University Press.
Corbett, J. and T. Jenkinson (1996), `The Financing of Industry, 1970±1989: An
International Comparison', Journal of the Japanese and International Economies, 10,
71±96.
ÐÐÐÐ (1997), `How is Investment Financed? A Study of Germany, Japan, The United
Kingdom and the United States', The Manchester School Supplement, 69±93.
Dale, S. and A. G. Haldane (1995), `Interest Rates and the Channels of Monetary
Transmission: Some Sectoral Estimates', European Economic Review, 39, 1611±1622.
Davis E. P. (1998), `Pension Fund Reforms and European Financial Markets', Financial
Markets Group, London School of Economics, Special Paper 107.
Deutsche Bundesbank (1995), `Verbriefungstendenzen im deutschen Finanzsystem und
ihre geldpolitische Bedeutung', Monatsbericht , April, 19±34.
Dewatripont, M. and E. Maskin (1995), `Credit and Ef®ciency in Centralized and
Decentralized Economies', Review of Economic Studies, 62, 541±55.
Dornbusch, R., C. Favero and F. Giavazzi (1998), `Immediate Challenges for the
European Central Bank', Economic Policy, 26, 15±64 (including discussion).
Edwards, J. and K. Fischer (1994), Banks, Finance and Investment in Germany, Cambridge,
Cambridge University Press.
Elsas, R. and J. P. Krahnen (1998), `Is Relationship Lending Special? Evidence from
Credit-File Data in Germany', Journal of Banking and Finance, 22, Special Issue on Credit
Risk Assessment and Relationship Lending, 1283±1316.
European Central Bank (1999), Possible Effects of EMU on the EU Banking Systems in the
Medium to Long Term, Frankfurt: European Central Bank.
Franks, J. and C. Mayer (1994), 'Corporate Control: A Comparison of Insider and
Outsider Systems', London Business School, Working Paper.
238 Reinhard H. Schmidt
Franks, J. and C. Mayer (1997): Corporate Ownership and Control in the United Kingdom,
Germany and France, Journal of Applied Corporate Finance, 9, 30±45.
Franks, J., C. Mayer and L. Renneboog (1998), 'Who Disciplines Bad Management',
London Business School, Working Paper.
Freixas, X. and J. C. Rochet (1997), Microeconomics of Banking , Cambridge, Mass: The
MIT Press.
Ganley, J. and C. Salmon (1997), 'The Industrial Impact of Monetary Policy Shocks:
Some Stylised Facts', Bank of England Working Paper 68.
Giovannetti, G. and R. Marimon (1998), 'An EMU with Different Transmission
Mechanisms', CEPR Working Paper 2016.
Goodhart, C. A. E. (1989), `The Conduct of Monetary Policy', Economic Journal, 99, 293±346.
ÐÐÐÐ (1993), `Banks and the Control of Corporations', The Cyprus Journal of
Economics, 6 (2), 91±108.
Goergen, M. and L. Renneboog (1998), Strong Managers and Passive Institutional Investors
in the UK, Louvain, European Corporate Governance Network.
Goux, J.-F. (1996), `Le canal etroit
du credit
en France', Revue d'economie politique, 104,
655±681.
Guender, A. and M. Moersch (1997), `On the Existence of a Credit Channel of Monetary
Policy in Germany', Kredit und Kapital, 30, 173±185.
Hackethal, A. (1999), Unternehmens®nanzierung in Deutschland, USA und Japan, PhD
dissertation, University of Frankfurt (fortcoming), Frankfurt, Lang.
Hackethal, A. and R. H. Schmidt (1999), 'Financing Patterns: Measurement Concepts
and Empirical Results', University of Frankfurt,Working Paper Series: Finance and
Accounting, 33.
ÐÐÐÐ (2000), Komplementarita t und Finanzsystem, Kredit und Kapital, Beiheft 15
(inprint).
Hackethal, A. and M. Tyrell (1998), 'Complementarity and Financial Systems ± A
Theoretical Approach', University of Frankfurt, Working Paper Series: Finance and
Accounting, 11.
Harhoff, D. and T. Korting (1998), `Lending Relationships in Germany ± Empirical
Evidence from Survey Data', Journal of Banking and Finance, 22, Special Issue on Credit
Risk Assessment and Relationship Lending, 1317±53.
Illing, G. (1997), Theorie der Geldpolitik: Eine spieltheoretische Einfuhrung, Berlin: Springer.
James, C. (1987), `Some Evidence on the Uniqueness of Bank Loans', Journal of Financial
Economics, 19, (2), 217±35.
Kashyap, A. K. and J. C. Stein (1997a), `The Role of Banks in Monetary Policy: A Survey
with Implications for the European Monetary Union', Economic Perspectives, Federal
Reserve Bank of Chicago, 22, (5), 2±18.
ÐÐÐÐ (1997b), 'What Do a Million Banks Have to Say About the Transmission of
Monetary Policy?', NBER Working Paper, 6056.
Kashyap, A. K. , R. Rajan and J. C. Stein (1999), 'Banks as Liquidity Providers: An Explanation
for the Co-Existence of Lending and Deposit-Taking', NBER Working Paper, 6962.
Kuppers, M. (1998a), 'Curtailing the Black Box: German Banking Groups in the
Transmission of Monetary Policy', University of Cologne, Staatswissenschaftliches
Seminar Discussion Paper Series, 17.
ÐÐÐÐ (1998b), 'On the Existence of a Credit Channel of Monetary Policy in
Germany: A Comment', University of Cologne, Staatswissenschaftliches Seminar
Discussion Paper Series 19.
La Porta, R., F. Lopez-de-Silvanes, A. Shleifer and R. W. Vishny (1999), 'Agency Problems
and Dividend Policies Around the World', Harvard University, Working Paper.
Differences Between Financial Systems in European Countries 239
Charles A. E. Goodhart
Schmidt's main result, as you will have read, is that ®nancial systems consist of
a number of complementary elements, with both Germany and the United
Kingdom having consistent but differing ± even polar opposite ± systems. I
accept this, and I am strongly predisposed towards accepting the counter-
attack launched against the Mayer position by Schmidt and his co-author and
student, Hackethal ± which is that the appropriate comparison is on gross, not
net, ®nancial ¯ows, and that on this basis the differentiation between the
German±Japanese and Anglo±Saxon ®nancial systems does show up reason-
ably clearly, (whereas it does not do so on a net basis).
I was also very interested in Schmidt's Figure 6.3 (p. 224) showing that the
French ®nancial system of intermediation was in several respects in the
process of moving from the German towards the British camp, as they had
presumably surmised in the Banque National de Paris (when launching a
hostile takeover bid). Indeed I had very little to criticise in sections on the
classi®cation of the various channels of monetary effects or on the general
characterisations of the ®nancial systems in these three countries.
But my role as a discussant is to comment critically and my particular
interest is on the effect of monetary policy on the real economies in Germany
and the UK. For this purpose I want to focus on Schmidt's Figure 6.4 (p. 228)
(see Figure D6.1).
Let me start with the credit channel. Here Schmidt claims that the effect on the
®nancial sector is larger in Germany than in the United Kingdom, largely because
the banks play a more central role in Germany, and I am prepared to accept that.
He also claims that the credit channel has a greater overall effect than the interest
rate channel and I again agree that that certainly can be the case, (consider, for
example, current Japanese problems), but I do not feel that this is necessarily so.
The credit channel, in my view, does not operate in a strictly linear fashion. In
some cases ± for example, where bank capital is already very plentiful ± it may be
weak, almost non-existent. In other circumstances it may, indeed, be ferocious.
So my classi®cation of the credit channel is as shown Figure D6.1b, distinguishing
between the weak and the (bracketed) strong cases.
241
242 Charles A. E. Goodhart
Credit channel
Germany
United Kingdom
Germany
United Kingdom
As revised by Goodhart
Credit channel
Germany
United Kingdom
Germany
United Kingdom
Figure D6.1 Monetary policy effectiveness in Germany and the United Kingdom
243
244
18
16
14
12
10 United Kingdom
Kingdom
Germany
8
0
1980:1 1982:1 1984:1 1986:1 1988:1 1990:1 1992:1 1994:1 1996:1 1998:1
United Kingdom
2 Kingdom
Germany
–0
–2
–4
–6
1980:1 1982:1 1984:1 1986:1 1980:1 1988:1 1990:1 1994:1 1996:1 1998:1
245
Table D6.1 A descriptive statistical analysis of nominal and real money market rates for
the United Kingdom and Germany, l980±l998
But the truth is that I do not have a good, cast-iron, explanation for my
quandary ± which is, to repeat, that, although I believe monetary policy bites
harder on output in the United Kingdom than in Germany, such policy has
nonetheless been more variable in the United Kingdom.
Let me conclude with a brief comment on dynamics. Because Schmidt sees
the current static position on the effectiveness of monetary policy in the
United Kingdom and Germany as balanced, he is more worried about dynamic
shifts in structure causing future problems. I do not share that worry. I worry
that doubts whether `one size ®ts all' in monetary policy in the Euro Zone
could be initially exacerbated by different ®nancial structures, with adverse
political repercussions. So I view a growing homogenisation of ®nancial
structures across Europe with more and longer ®xed-term mortgages in the
United Kingdom and more reliance on securities markets in Germany as an
almost unalloyed bene®t.
Discussion
Alain Vienney
247
248 Alain Vienney
levels, underplays the interest rate channel, which is by far the predominant
one and explains a number of differences between countries. To judge by the
number of stars allocated in Figure 6.4 (p. 228), Schmidt regards the credit
channel as even more important than the interest rate channel, which comes
as a surprise.
between the impact of monetary policy on the ®nancial economy (that is the
in¯uence of movements in key rates on bank lending rates and on short- and
long-term interest rates) and the impact of the ®nancial economy on the real
economy.
. The net interest margin may allow some leeway on the bank lending rates
and it still differs widely in the Euro Area
. The presence of regulated rates is of primary importance for the setting up
of lending rates, in France, one-third of M3 is made up of this kind of assets
with administrated rates and it is part of the public debate to argue that
these rates prevent an additional lowering of the borrowing terms.
Simulations conducted, under the aegis of the BIS, by the central banks using
their national macroeconomic models and collated by Smets9 demonstrate a
greater responsiveness to a `standardised' monetary shock in the case of the UK
economy than in the case of France or Germany. Thus, a one-point increase in
the key rate over two years leads in the very ®rst year to a 0.85 per cent
contraction in GDP in the United Kingdom, compared with more limited
declines of 0.4 per cent and 0.37 per cent, respectively, in France and
Germany.10
While the VAR models11 produce more contrasting results, they also
highlight differences in the scale of the adjustment that economies make to a
monetary shock, although the speed of adjustment is more or less the same
250 Alain Vienney
from one country to another (in particular, no price effect is evident within a
period of under eighteen months±two years).
It will be noted, however, that the results for the United Kingdom are very
close to those obtained for Italy and that, in this respect also, there is no
essential difference between the United Kingdom on the one hand and
continental Europe on the other.12
Notes
1. In 1995, the net position in interest-bearing assets were 48 per cent, for German
households, 26 per cent for France and ±9 per cent for the United Kingdom.
2. As a general rule, the United Kingdom prefers variable or adjustable rates (for
example, in the case of mortgage lending), whereas most loans and assets in
Germany and France are still at ®xed rates.
Discussion 251
References
Barrah F., V. Coudert and B. Mojon (1994) `Transmission de la politique mone taire et
credit bancaire, une application a cinq pays del', OCDE CEPII Morning Paper, 94±03,
June.
Borio, C. E. V. and L. S. Fitz (1995) `The Response of Short-term Lending Rates to Policy
Rates: A Cross-country Perspective', Basle, BIS, May.
Browne, F. and L. Fell (1996) `The Changing Institutional Context of Monetary Policy
and its Implication for In¯ation', Frankfurt, EMI, May.
Dornbusch, R., C. Favero and F. Giavazzi (1998) `A Red Letter day', CEPR Discussion Paper,
1804.
Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from
the G7 Countries', Basle, BIS.
P®ster, C. and T. Grunspan (1998) `Bank Restructuring, Monetary Policy Ef®ciency and
Financial Stability in France', Central Banks' Economists' Metings, Basle, 28±29
October.
Smets, F. (1995) `Central Bank Macroeconomic Models and the Monetary Policy
Transmission Mechanism', Basle, BIS.
7
1 Introduction
252
European Labour Markets and the Euro 253
discussion one step further: for a number of reasons, the arrival of EMU will
itself have signi®cant effects on the functioning of labour and product markets
and the relative importance of real and nominal rigidities. These feedbacks will
ultimately affect the way Europe reacts as a macroeconomic entity to demand
disturbances and how its central bank views the effectiveness of monetary
policy.
This chapter surveys a number of issues too involved to be treated in model-
theoretic detail here. I will furthermore abstain from econometric analyses for
reasons which should be clear to all. There is a sense that the macroeconomic
regime has changed in a way it has not in several hundred years in Europe: if
the Lucas Critique has any relevance at all, it had better be here and now. I will
adduce some empirical evidence however, which is suggestive of what one
might expect in the future. The chapter is highly speculative, but meant to be
so.
My discussion is organised as follows. In Section 2, I discuss the
macroeconomic impact ± at both regional and pan-European level ± of the
current structure of labour markets. Section 2 surveys the multifarious means
by which a monetary union could affect the functioning of labour markets.
This feedback takes some surprising turns, and may lead to a wholly different
perception of the transmission channels of monetary policy in Europe. Section
3 adduces simple but striking evidence in support of my hypotheses and
Section 4 concludes.
Note:
a
Numbers in parentheses are described as highly unreliable.
Figure 7.1 Complementarity of real and nominal rigidities for a given price change
prices, given that others are not doing so. Strategic complementarity implies
that second-order issues for the ®rm can have ®rst-order effects for the
macroeconomy.
Money wage rigidity could also induce business cycle ¯uctuations. While an
important element in the early intellectual development of Keynesian
macroeconomics, nominal wage rigidity is not borne out at the micro level
(Bils, 1985, Smith, 1999) nor is it particularly supported by aggregate evidence
on wage and price dynamics (see references in Blanchard, 1990); Jeanne (1998)
and Roger (1998) have both shown that nominal price rigidity, combined with
some degree of real wage rigidity, is suf®cient to generate persistent
¯uctuations that resemble US business cycles.10
Summary
The previous discussion can lead to rather sombre conclusions about the
future of EMU. First, the conventional wisdom of extreme rigidity in labour
markets, which now has the OECD seal of approval (OECD, 1994) and is
accepted nowadays by everyone except the labour unions and perhaps a few
surviving extremists in the German ®nance ministry, should render the EMU a
Mundellian nightmare. It won't be necessary, according to this logic, for
another German reuni®cation to occur to generate real problems. All we need
is some overheating in Ireland, Portugal, or Finland and the whole EMU
project will collapse as the other regions slump without any equilibrium
mechanism.
An equally pessimistic message emerges on the monetary transmission
mechanism when considered under these circumstances, in which a rapid
pass-through into in¯ation is taken for granted by market participants.
Reviews by Buti and Sapir (1998) and Dohse and Krieger-Boden (1998) give
rather sombre pictures of the prospects, and Dornbusch, Favero and Giavazzi
(1998) raise questions about the asymmetric impact of monetary policy on the
participating EMU countries. Moreover, ®scal policy is hamstrung by the
Maastricht Treaty and the Pact for Stability and Employment and potential
exists for `beggar-thy-neighbour' effects as countries jockey to better their
macroeconomic circumstances. This `Flassbeck±Lafontaine Hypothesis' sees
purposeful competitive de¯ation just around the corner, as countries unable to
devalue are forced to regain competitiveness by more painful means. In this
view, governments, robbed of their power to generate instant nominal
devaluations will do what Britain did in the ®rst half of the 1920s. Feldstein's
(1992) criticism is now widely accepted that politics have outweighed
economics; Eichengreen (1998) has already speculated about the `dissolution'
of the European Monetary Union before it even begins.
Given this doomsday scenario, critical economists are compelled to ask the
question: Are rigidities in Europe set in stone? Is it reasonable to assume that
the Euro will leave labour markets and their institutions intact ± and, if not,
which ones are implicated? What will be the consequences of these changes?
258 Michael C. Burda
currency, for reasons stressed by Akerlof and Yellen (1985), Mankiw (1985)
and Romer (1996). Increased exposure to the sheltered domestic market will
mean greater incentives to price to market and to set nominal prices in
advance for longer periods, as customer relations become more important and
the net bene®ts of charging stable nominal prices increase (Okun, 1981).
The third and potentially most important effect ¯ows from the credibility
that comes from having a central bank which can `stand above' (i.e. ignore)
economic conditions in individual countries and be free of political pressure.
To this extent if the European Central Bank (ECB) is really the most
independent central bank in the world, agents will be more prone to expect
low in¯ation and will not attribute deviations to policy changes. This
important source of inertia should be distinguished from the usual wage±
price mechanism (e.g. Blanchard, 1990); rather it has to do with the anchoring
of in¯ationary expectations and the effect this will have on the willingness to
negotiate contracts in nominal terms.
To give some sense on the evolution of rigidities, I present some simple
statistics for data on comparable price and wage time series from EU member
countries.14 Table 7.2 displays average unweighted correlations of bilateral
in¯ation rates (®rst difference in the logarithms) for a number of groupings of
countries in addition to the Euro-11 since 1961. For comparison, I present data
for eight regions of the United States for a similar time period. Clearly, an
increase in price convergence has taken place across the board, not only in the
smaller `core' groupings. The eigenvalues of the moment matrix indicates the
extent to which in¯ation in one country can be expressed as a linear
combination in others. Table 7.3 documents that, to a large extent, my
Figure 7.2 The cost of passive quantity adjustment in response to an exchange rate
depreciation
260 Michael C. Burda
Table 7.2 Synchronisation of price in¯ationa in Europe and the United States, 1961±96
Core Europe 0.76 0.80 0.82 9.82 10�4 5.89 10�4 ±39.9
(B, NL, D, A) (0.08) (0.06) (0.09) 0.207 0.0928 ±55.2
Core Europe 0.74 0.71 0.81 8.74 10�4 1.64 10�4 ±81.2
F, DK, IT (0.11) (0.13) (0.12) 0.560 0.363 ±35.1
Euro-11 lightb 0.73 0.73 0.80 6.53 10�4 3.88 10�5 ±94.1
(0.14) (0.14) (0.15) 0.983 0.602 ±38.8
Euro-11 lightb 0.71 0.70 0.78 3.17 10�4 3.69 10�5 ±88.4
DK, S, UK (0.13) (0.15) (0.14) 1.336 0.811 ±39.3
Memo: USA
8 regions, 0.95 _ _ 1.48 10�5 _ _
1978±92, (0.03) 0.376
GSP de¯ator
Notes: a In¯ation is measured as ®rst difference in the logarithm of the relevant price index.
b
Less Luxembourg, Portugal.
conclusions hold when looking at a much smaller time interval and when
correcting for exchange rate changes.
It has been argued, by Calmfors (1998a) and others, that monetary union
could result in increasing nominal money wage rigidity. Presumably this
would arise as a result of the low level of in¯ation and resistance to nominal
wage reductions. In addition, the alignment of traded goods prices should
impose factor price convergence, as long as complete specialisation does not
occur ®rst, although this can only be a statement about labour of a given
quality. At the same time, Calmfors (1998a) claims that increasingly variable
macroeconomic conditions might lead to shorter nominal contract periods
and greater nominal wage ¯exibility.
Nominal wage behaviour in Europe over the past thirty years lends support
to my contention that nominal wages are less likely to be rigid than prices.
Table 7.4 and 7.5 clearly show a determination in the strong positive
correlation of real wage growth present in the 1960s and 1970s. To the extent
European Labour Markets and the Euro 261
Source: OECD.
Table 7.4 Synchronisation of nominal wage growtha in Europe and United States,
1961±96
Core Europe 0.85 0.76 0.46 2.15 10�3 1.24 10�3 ±42.3
(B, NL, D, A) (0.06) (0.17) (0.11) 0.688 0.130 -81.0
Core Europe 0.72 0.52 0.48 1.49 10�3 6.59 10�4 ±55.8
F, DK, IT (0.15) (0.32) (0.18) 1.49 0.451 ±69.8
Euro-11 lightb 0.71 0.46 0.55 5.58 10�4 2.57 10�4 ±54.0
(0.15) (0.35) (0.22) 2.39 0.760 ±68.2
Euro-11 lightb 0.66 0.48 0.50 1.80 10�4 6.07 10�5 ±66.2
DK, S, UK (0.18) (0.31) (0.26) 2.96 0.981 ±66.9
Memo: USA
8 regions, 0.92 _ _ 2.01 10�5 _ _
1978±92, (0.06) 0.449
annual comp.
Memo: USA
8 regions, 0.90 _ _ 1.65 10�5 _ _
1978±92, (0.08) 0.425
wages/salaries
Notes: a Nominal wage growth is measured as ®rst difference in the logarithm of the wage index.
b
less Luxembourg, Portugal.
Sources: US: Bureau of Economic Analysis (REIS), IMF, International Monetary Statistics.
Table 7.5 Nominal manufacturing wage growth correlations in national currency and
DM termsa
Notes: a First differences in log hourly nominal compensation costs for production workers in
Table 7.6 Nominal wages in manufacturing in the EU, 1986 and 1996
Summary
What are the macroeconomic implications of increasing nominal rigidity and
real ¯exibility, ceteris paribus? The empirical evidence, which is meant to be
suggestive, supports the contention that nominal price rigidity has increased
as a consequence of product market integration and exchange rate stability.
Nominal wages, in contrast, are highly correlated only in the core, and this
applies a fortiori to real wages and real exchange rates as well. These ®ndings
suggest that the Euro will affect labour market ¯exibility in the direction of
more ef®ciency. Without more detailed information on preferences, it is
impossible to say whether this increase in ef®ciency will lead to overall welfare
gains; some analyses, such as Agell (1998), claim that labour market rigidities
may re¯ect welfare-improving policies in the light of other market imperfec-
tions. Burda (1995) has presented a related rationale for union wage
compression.
European Labour Markets and the Euro 267
Table 7.7 Synchronisation of real wage growtha in Europe and the United States,
1961±96
Core Europe 0.60 0.69 0.24 2.69 10�3 9.04 10�4 ±66.4
(B, NL, D, A) (0.08) (0.16) (0.38) 0.170 0.014 ±91.5
Core Europe 0.59 0.45 0.08 1.76 10�3 2.76 10�4 ±84.3
F, DK, IT (0.13) (0.24) (0.41) 0.291 0.018 ±93.6
Euro-11 lightb 0.55 0.36 0.06 9.96 10�4 1.937 10�4 ±80.6
(0.13) (0.25) (0.42) 0.405 0.026 ±93.5
Euro-11 lightb 0.46 0.35 0.14 5.62 10�4 1.35 10�5 ±97.6
DK, S, UK (0.20) (0.24) (0.39) 0.455 0.036 ±92.1
Memo: USA
8 regions, 0.59 _ _ 6.68 10�5 _ _
1978±92, (0.18) 0.016
real comp.)
US (8 regions,
salaries)
Notes: a Real wage growth is measured as ®rst difference in the logarithm of the nominal wage index
reported by the IMF, International Finance Statistics, divided by the IMF/IFS consumer price index.
b
Less Luxembourg, Portugal.
Concluding remarks
In addition to its historic dimensions, EMU will shed new light on a number of
old, bothersome questions. Naturally, it will help us understand better how
monetary unions function. In the ®rst instance, however, it will teach
economists and policy makers the relevance of the new-Keynesian approach
to understanding aggregate ¯uctuations, for which there is precious little
evidence. It will also help us decide whether nominal price or wage rigidities
are more relevant for explaining the real effects of aggregate demand
¯uctuations and thus the transmission mechanism itself.
The convergence of exchange rate and especially price dynamics suggests
that the preconditions for nominal price rigidities have become more
European Labour Markets and the Euro 269
Table 7.8 Manufacturing real wage growth correlations using different price indexes
Wages de¯ated by OECD price index Wages de¯ated by IMF price indexa
Per cent
14
United States
12
Germany
10 Netherlands
Changes to previous year
–4
–6
Figure 7.3 Growth rates of real hourly compensation costs in manufacturing, 1960±96
in the short run. If I turn out to be wrong and have to eat my hat, this fact will
nevertheless have been useful information for our profession as well as policy
makers. If I am right, EMU will have delivered the ultimate bonus in real
ef®ciency gains for the unemployment-riddled labour markets of the
continent.
Appendix
Notes
1. The view that short-run adjustment costs associated with EMU are small relative to
long-run gains has been echoed by Buiter (1995).
2. See Mundell (1961), as well as McKinnon (1963); Kenen (1969).
3. Indirect estimates of labour mobility for the United States by Blanchard and Katz
(1992) and for Europe by Decressin and Fatas (1995) show that European regions
tend to adjust to adverse employment shocks via changes in labour force
participation as opposed to residence. For more detailed summaries of the
evidence, see Eichengreen (1993) and Gros and Hefeker (1998) as well as Obstfeld
and Peri (1998).
4. See Gros and Hefeker (1998) for an overview.
5. It is remarkable that the optimal currency literature has largely ignored the role of
capital mobility ± meaning long run mobility of the means of production ± despite
Mundell's own explicit reference to it in his seminal article. For examples, see
discussions in Bo®nger (1994); Bayoumi and Eichengreen (1996); Wyplosz (1997);
Gros and Hefeker (1998).
6. See Bruno and Sachs (1985); Sachs (1979, 1983); but also Branson and Rotemberg
(1980).
7. Among others, Romer (1996) has stressed the labour market as a primary source of
real rigidities in the macroeconomy, as complementary to nominal rigidities.
272 Michael C. Burda
8. `The argument for ¯exible exchange rates is, strange to say, very nearly identical
with the argument for daylight savings time. Isn't it absurd to change the clock in
summer when exactly the same result could be achieved by having each individual
change his habits? All that is required is that everyone decide to come to his of®ce
an hour earlier, have lunch an hour earlier, etc. But obviously it is much simpler to
change the clock that guides all than to have each individual separately change his
pattern of reaction to the clock, even though all want to do so.' (Friedman (1953),
p. 173, my emphasis).
9. In fact, the ®rm in panel (b) is more likely to ration output, producing only to the
point at which price equals marginal cost, and thereby violating the assumption of
completely passive (i.e. demand-determined) adjustment of production to demand.
In any case the point is made that incentives to change prices in this case are large.
10. For evidence on the rigidity of prices in the United States see Carleton (1986),
summaries of empirical evidence are available in Blanchard (1990) and Romer
(1996).
11. This argument can also be found in McKinnon (1963), who stresses the role of non-
traded goods in the reaction to devaluations.
12. The failure of ®rms selling into the United States fully to pass through exchange rate
¯uctuations is well documented (see Knetter, 1989; Dornbusch, 1987, 1996) and
could be seen as an indication of what Euroland can expect.
13. One exception could be energy prices, which continue to be denominated in
dollars. As Europe is the largest customer of the oil-exporting Middle East and
Russia it may come to pass that oil prices are denominated in Euros. The relevant
issue, of course, is whether oil prices in Euros will tend to become more stable over
time.
14. The empirical evidence I present in this chapter is rather modest, as it seems foolish to
place much weight on estimates of structures in place before monetary union. On the
other hand many investigators have looked at the temporal evolution of cross-
correlation of price and quantity variables. (For example DeGrauwe, 1991; von Hagen
and Neumann, 1994; Bayoumi and Eichengreen, 1996; Frankel and Rose, 1996). For
details on the data used, the reader is referred to the Appendix (p. 271).
15. The Marshall±Hicks rule states that the elasticity of demand for labour is higher, the
higher the elasticity of demand for output produced with that labour, the higher
the elasticity of substitution between labour and other inputs the lower the
elasticity of supply for those competing inputs and the greater the cost share of
labour in production. See Hamermesh (1993).
16. For evidence on how product market competition has affected labour unions and
labour markets in the United States in general, see Duca (1998).
17. Calmfors (1998a, 1998b); Gru È ner and Hefeker (1998); Cukierman and Lippi (1998,
1999); Soskice and Iversen (1999).
18. Arguing from a Barro±Gordon perspective, Calmfors (1998c) has conjectured that
incentives to reform inside the EMU are greater than outside, since countries with
control over monetary policy are likely to view labour market reform and monetary
policy as substitutes for reducing unemployment, while inside EMU the latter
vanishes. Reforming labour markets provides one means of ensuring against
idiosyncratic shocks. This effect will be strengthened by ®scal pressures stemming
from increasing unemployment, as well as the reorientation of national objective
functions when in¯ation can no longer be in¯uenced by national policies.
Similarly, Hefeker (1998) assumes unions which choose both the nominal wage
and the degree of ¯exibility.
European Labour Markets and the Euro 273
References
Agell, J. (1998) `On the Bene®ts from Rigid Labour markets: Norms, Market Failures, and
Social Insurance', Uppsala Dept of Economics, Working Paper, 1998:17.
Akerlof, G. and J. L. Yellen (1985) `A Near-rational Model of the Business Cycle, with
Wage and Price Inertia', Quarterly Journal of Economics, 100, 823±38.
Ball, L. and D. Romer (1990) `Real Rigidities and the non-Neutrality of Money', Review of
Economics and Statistics, 57, 183±203.
Barro, R. and D. Gordon (1983) `A Positive Theory of Monetary Policy in a Natural Rate
Model', Journal of Political Economy, 91, 589±610.
Bayoumi, T. and B. Eichengreen (1993) `Shocking Aspects of European Monetary
Integration', in F. Torres F. and F. Giavazzi, Adjustment and Growth in the European
Monetary Union, Cambridge, Cambridge University Press.
ÐÐÐÐ (1996) `Operationalizing the Theory of Optimum Currency Areas', CEPR
Discussion Paper, 1484.
Bean, C. (1998) `The Interaction of Aggregate Demand Policies and Labor Market
Reform', Swedish Economic Policy Review 5, 353±82.
Belke, A. (1996) `Implikationen einer zentralisierten Geld- und WaÈhrungspolitik fuÈ r die
BeschaÈftigung in Europa', 8 UniversitaÈt Bochum, Diskussionspaper.
Berthold, N. and R. Fehn (1998) `Does EMU Promote Labor-Market Reforms?', Kyklos, 51,
509±36.
Bils, M. (1985) `Real Wages and the Business Cycle: Evidence from Panel Data', Journal
of Political Economy, 93, 666±89.
Blanchard, O. (1990) `Why does Money Affect Output: A Survey', Chapter 15 in
B. Friedman B. and F. Hahn (eds), Handbook of Monetary Economics, Amsterdam, Elsevier.
Blanchard, O. and L. Katz (1992) `Regional Evolutions', Brookings Papers on Economic
Activity, 1, 1±75.
Bo®nger, P. (1994) `Is Europe an Optimal Currency Area?', CEPR Discussion Paper, 915.
Branson, W. and J. Rotemberg (1980) `International Adjustment with Wage Rigidity',
European Economic Review, 13 (3), 750±77.
Bruno, M. and J. Sachs (1985) Economics of Worldwide Stag¯ation, Cambridge, MA,
Harvard University Press.
Buiter, W. (1995) `Macroeconomic Policy during a Transition to Monetary Union', CEPR
Discussion Paper, 1222.
Burda, M. (1995) `Unions and Wage Insurance', CEPR Discussion Paper, 1232.
Buti, M. and A. Sapir (1998) Economic Policy in the EMU, Oxford: Oxford University Press.
Calmfors, L. (1998a) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference does EMU Make?', Oxford Review of Economic Policy, 14, 125±51.
ÐÐÐÐ (1998b) `Unemployment, Labour-market Reform and Monetary Union',
Institute for International Economics Studies Seminar Paper, 639.
ÐÐÐÐ (1998c) `Monetary Union and Precautionary Labour-market Reform', Institute
for International Economics Studies Seminar Paper, 659.
Calmfors, L. and, Drif®ll (1988) `Bargaining Structure, Corporatism and Macroeconomic
Performance', Economic Policy, 6, 3±62.
Carleton, D. (1986) `The Rigidity of Prices', American Economic Review, 76, 637±58.
274 Michael C. Burda
Hendrik J. Brouwer
276
Discussion 277
Burda tells us that the call for political action to achieve greater ¯exibility does
not really matter much, since EMU itself will force labour markets in Europe to
become suf®ciently more competitive. His policy message is surprising and
indeed provocative. The question is whether that is the whole story. Let me
®rst note the issues on which I agree with Michael Burda:
. First, the chapter's central notion that EMU implies a major regime shift
which will not let the existing labour institutions in Europe remain intact.
. I also agree with the expectation that EMU may increase nominal price
rigidities. Not only the reasons mentioned in the chapter ± a reduced role
for exchange rate changes and for outside competition (pp. 258±61) ± could
play a role, but also the independent European Central Bank (ECB) with a
mandate of maintaining price stability will make a stable price climate in
Europe more likely. The general view that ®scal policy should contribute to
stability ± as expressed in the Stability and Growth Pact ± gives further
support.
. I share Burda's scepticism about the likelihood of centralised wage
negotiations in Europe (pp. 264±5). But there are some tendencies in that
direction. Therefore, I would like to go one step further: Is it really useful to
propagate and stimulate a European Social Dialogue? I doubt whether that
could enforce labour market ¯exibility. So far, the results of this dialogue
have produced the opposite.
. The chapter states that real rigidities will diminish through more
competition in the labour market which will lead to weakening of union
power in wage determination (p. 264). This contrasts with the recent high
wage demands in Germany, which do not give the impression of
moderation on the side of the unions brought forth by EMU.
. Of course, one could argue that such high wage demands are transitory but
that still leaves the question of what time frame Burda has in mind with
regard to this change in rigidities. In the meantime, when moving from one
state of equilibrium to another, things could still go wrong. Hysteresis
effects explain why Europe has such a high level of structural
unemployment.
. What could go wrong is not only that wage demands are excessive, but also
that governments resist the transfer of a part of their coordination role to
the market. On the contrary, painful adjustment in the labour market may
well lead to a call for increased coordination by governments in the
direction of short-term `quick ®xes', like the 35-hour working week, while
neglecting the need for sometimes unpopular structural adjustment.
278 Hendrik J. Brouwer
1 Introduction
Burda's chapter 4 is not an easy chapter to comment on. Its aim is `to entertain
new-Keynesian arguments' without a well speci®ed model and, therefore, no
clear policy recommendation can be derived from it, despite being related to
policy issues. Furthermore, it is also not an empirical paper, because `the
empirical evidence is rather modest' and it is questionable that such evidence
would be of any relevance owing to the change of regime, as correctly
emphasised by Burda. The chapter deals with institutional and sociological
matters for discussing the impact of the change of economic regime due to
monetary union, with particular emphasis on labour markets. As Burda says,
`is highly speculative, but meant to be so'. I would say it is both speculative and
provocative, and this is an excellent idea.
Given these dif®culties, let me summarise my understanding of the main
issues raised by the chapter and make some comments as I go along.
Some of the issues the chapter promises to address are those found in the
literature on optimal currency areas (OCA) following Mundell (1961). What is
the optimal response of monetary policy to idiosyncratic shocks in an
economy? Should there be one or multiple currencies, in order to respond
optimally to these shocks? As the literature stands now, we are not in a
condition to answer all these questions satisfactorily. However, one thing is
certain: the Mundellian nightmare regarding OCA in the context of EMU has
been oversold. The understanding of exchange rate policy is very different
today than it was in 1961. More speci®cally, it is today widely accepted that no
exchange rate policy can insulate an economy from real shocks. To assert that
exchange rates can do more than insulate a country from nominal shocks is an
incorrect reading of Mundell (1961). This clari®cation has been recently made
in Mundell (1997, 1998).
279
280 LuõÂs Campos e Cunha
Price rigidities
The chapter also argues that as the regime changes, price stickiness in the
goods market will increase. The author claims that EMU will lead to an
increase in nominal rigidities because the share of `home goods', goods traded
inside Europe, is larger than the total of home goods across all countries in the
Union. The reasoning is that these goods will not be exposed to ¯uctuations
and this will favour nominal rigidities. Both parts of the argument need
clari®cation: is Burda arguing that shocks to goods traded inside the EU were
mainly driven by national monetary policies? Otherwise, ¯uctuations in these
sectors will certainly persist. Furthermore, why should a more stable
environment lead to higher nominal rigidities?
This `inwardisation' effect owing to the single currency is probably too
much emphasised, since most of the effects were anticipated at least for the so-
called `core' countries. Furthermore, the `closing' of the Euroland economies
does not have the same nature of a country imposing tariff barriers and capital
controls. With the single currency the domestic economies become bigger,
and there is an increase in competition owing to reduced transaction costs and
increased price transparency. Market segmentation is less likely to take place
and, therefore, ®rms are more ± not less ± exposed to competition as economic
frontiers are pushed further away, so it is not obvious that monopolistic power
will necessarily increase.
I would like to believe, using the wording of the chapter, that, `EMU is the
Trojan horse of decentralisation' and that the `Euro is the Trojan horse which
will liberalise labour markets', and therefore, `calls for additional ¯exibility
may be the economic equivalent of whipping a dead horse'. Unfortunately,
most of us feel that we have to continue to call for some structural reforms
leading to institutional and legal changes so that structural unemployment
will fall. The Euro will not be the solution for high unemployment rates.
Unlike Burda, I am more concerned with the symmetric consequences in the
new regime. We could imagine consequences concerning labour markets
owing to compulsory centralised harmonisation. In fact, regions of Euroland
with lower labour productivity should be vigilant and oppose attempts to
harmonise labour market regulation throughout the Union. Just imagine the
impact of a forceful and a speedy harmonisation of minimum wage towards
the core country levels. It would be the equivalent of Brussels offering a rope to
thousands of workers to hang themselves and it would be very dif®cult for
them to resist. This would imply a jump in unemployment rate for lower-
productivity regions as real rigidities increased. This is foreseeable and it would
be a big step in the wrong direction.
As regards monetary policy, Burda makes a point that stronger price
persistence in Euroland will facilitate a more powerful monetary policy, which
282 LuõÂs Campos e Cunha
References
Lucas, R. (1976), `Econometric Policy Evaluation: A critique', in K. Brunner and
A. H. Meltzer (eds), The Phillips Curve and Labor Markets, Amsterdam, North-Holland,
19±46.
Mundell, R. (1961), `The Theory of Optimal Currency Areas', American Economic Review,
51, 657±65.
ÐÐÐÐ (1997) `Currency Areas, Common Currencies and EMU', American Economic
Review, (2) 214±16.
ÐÐÐÐ (1998) `Great Expectation for the Euro', The Wall Street Journal, March, 24±5.
8
1 Introduction
283
284 Otmar Issing
There is by now quite a large literature on the MTP in the Euro Area countries.
While the focus of this literature has often been on the cross-country
differences and the implications for the common monetary policy, a lot can
also be learned from this literature concerning Euro Area-wide transmission.
Two approaches have been used which I will discuss in turn. One approach is
to describe and analyse the institutional features of the economies which have
a bearing on the MTP. Another is to use econometric models to estimate the
effects of a change in policy-controlled interest rates on economic activity and
in¯ation. Let me brie¯y review the results of both approaches.
Institutional comparison
The way in which changes in monetary policy feed through the various sectors
of the economy depends to a large extent on the structural and institutional
features of its ®nancial, labour and goods markets. These institutions are
rooted in the history of each country and often resulted from the interplay
between particular macroeconomic shocks, the preferences and bargaining
powers of various agents in the economy and more or less market-friendly
regulatory and government forces.
Researchers have devoted a lot of energy to investigating whether such
institutional characteristics, which from a theoretical perspective may affect
the impact of monetary policy, differ across countries or regions. Often the
focus on one particular aspect of the MTP leads to the conclusion that the
transmission of monetary policy impulses will be very heterogeneous across
Euro Area countries.
However, if one brings several of these features together (as is done in
Table C1.1), then it is much less obvious that the effects of policy will be very
different. Generally, countries that are particularly sensitive to policy changes
because of one criterion, will be less sensitive on the basis of other
characteristics. Let me brie¯y go through some examples.
Goods markets
One important characteristic of the goods markets which affects the transmis-
sion channels of monetary policy is the degree of openness. While the Euro
Area as a whole will be a relatively closed economy ± with a ratio of exports to
Concluding Remarks 285
GDP of about 15 per cent, it is only marginally more open than the United
States or Japan ± there is a fair degree of dispersion within the Euro Area (see
Table C1.1). For example, Belgium is about three times as open to the non-
Euro Area as Spain. As a consequence, while the exchange rate channel may be
relatively less important for the Euro Area as a whole, the strength of its impact
on output and in¯ation may be quite different across countries.
Similarly, the overall interest elasticity of aggregate demand will depend on
the importance of interest rate sensitive sectors in the economy. For example,
Carlino and DeFina (1998) ®nd that, within the United States, regions with a
higher weight of construction and manufacturing in the economy are more
strongly affected by the Federal Reserve System's monetary policy. If similar
considerations are true within the Euro Area, then one could expect that, for
example, Germany and Spain should experience a stronger interest rate
channel than the Netherlands or Belgium.
Another sector which is highly responsive to the level of the interest rate
is the housing sector. Maclennan, Muellbauer and Stephens (1998) have
expressed concern about the implications of cross-country differences in the
tax and legal framework of housing markets for the transmission mechanism.
In particular, transaction costs owing to taxes on new houses and stamp duties
are quite heterogeneous across countries. A particularly low tax rate in the
United Kingdom is consistent with the high amplitude of its housing market
cycle. Within the Euro Area, Germany, Italy and Spain have much smaller
transaction costs than the Netherlands, France and Belgium.
Financial structure
Table C1.1 also summarises a number of features of ®nancial structure across
countries. From a comparison of these features across countries, a number of
tentative conclusions can be drawn. Countries in the Euro Area appear to have
several features in common. Stock markets and corporate debt securities
markets are less well developed than in the United States, the United Kingdom
or even in Japan. Of the seven largest Euro countries, the Netherlands is the
only one where the number of publicly traded ®rms and the stock market
capitalisation show levels that are comparable to those of the United States or
the United Kingdom. Obviously, this implies that equity prices in the Euro
Area should play a less important role in the MTP through both the effect on
the cost of capital and wealth effects on consumption.
Similarly, France is the only country where the corporate sector issues a
signi®cant share of its debt directly on the market. In 1997, up to one-third of
the debt of French ®rms consisted of securities, while it was below 10 per cent
in the other countries. To the extent that bank lending rates are more sticky
than market rates in response to a change in policy-controlled rates, the
interest rate channel should be weaker in the Euro Area.2
However, there remain some important differences in ®nancial structure
within the Euro Area. For example, the degree of ®nancial intermediation is
Table C1.1 Institutional features with a bearing on the transmission mechanisma
Non-Euro Manufactur. Stamp duty Number of Stock Securities Credit at Union Employment Unemploy.
exports and VAT publicly market / Total ®rm adjustable coverage protection bene®t
/ GDP construction on new traded ®rms capitalisation debt rate index index index
(per cent) / GDP homes / per capita / GDP (per cent) / GDP
(per cent) (per cent) (per cent) (per cent)
Source IMF, Carlino and Maclennan, Cecchetti Prati and Borio Borio Nickell Nickell Nickell
Direction of DeFina Muellbauer and (1999) Shinasi (1995) (1995) (1997) (1997) (1997)
Trade (1998) Stephens (1997)
Statistics (1998)
286
Concluding Remarks 287
quite different among the large Euro Area economies with loans of credit
institutions to non-banks ranging from as low as 67 per cent of GDP in Italy, to
as high as about 150 per cent of GDP in Germany and in the Netherlands.
Similarly, the maturity structure of the debt of the private sector differs
considerably across countries (Borio, 1995).
For the Eurosystem it will be even more important than for most other central
banks to continuously monitor any developments that may change the
transmission process of monetary policy. I would like to focus on the likely
developments in the ®nancial sector and its impact on the MTP.
I will not elaborate on the potential effects of the single currency on goods
and labour markets. Let me just underline that, particularly in the case of
labour markets, further structural reform is needed not only to reduce the
Table C1.2 Simulation of a monetary policy tightening
Monticelli and Tristani (1998) Peersman and Smets (1999) ECB area-wide model
Horizon Year 1 Year 2 Year 3 Year 1 Year 2 Year 3 Year 1 Year 2 Year 3
Interest rate 0.102 0.044 ±0.025 0.5 0.5 0 0.5 0.5 0.5
GDP ±0.18 ±0.4 ±0.38 ±0.18 ±0.32 ±0.27 ±0.07 ±0.16 ±0.23
Consumer prices 0.07 ±0.16 ±0.45 ±0.11 ±0.7 ±1.8 ±0.11 ±0.22 ±0.25
289
commercial paper and corporate bonds for non-®nancial ®rms may increase in
countries where it has not yet developed and reach levels that are currently
observed only in France (a third of the corporate sector debt). The potential
consequences of the development of direct ®nance and a changing role of the
banking sector in Europe for the transmission may be signi®cant. Competition
between direct ®nance and bank ®nance will put pressure on banks and may
increase the speed of transmission from the money market rate to the rate at
which ®rms borrow. For example, in France, evidence shows that the money
market rate has taken over the prime rate as the reference to compute the bank
credit rate to big ®rms as well as to smaller ones. In the short run, it may,
however, also increase the fragility of the banking sector.
On the other hand, cultural preferences for ®xed rate contracts and against
indebtedness, national consumer protection laws and inertia in consumer
habits will tend to slow down cross-country competition in retail banking.
Retail banking involves long-term investment in brand names, legal expertise
in speci®c national laws and monitoring of borrowers' solvency. These provide
incumbent banks with substantial advantages over competitors, especially
foreign ones, so that initially the impact of EMU on retail banking could be
limited to the harmonisation of banks' funding costs.
4 Conclusion
This review of the existing literature on the MTP in the Euro Area con®rms the
validity of Friedman's famous dictum that monetary policy lags are long and
variable ± and, I would add, very uncertain. In the case of the Euro Area,
uncertainties are magni®ed by the scarcity of reliable aggregate data and the
potential effects of a regime change. The monetary policy strategy of the
Eurosystem was designed to face, in a pragmatic, yet conceptually sound way,
the many uncertainties inherent in this regime change.
The strategy is at the same time transparent and ¯exible. It clearly
communicates the commitment to price stability by de®ning price stability
as an increase of the area-wide harmonised index of consumer prices of below
2 per cent. This helps to anchor in¯ation expectations and preserve the anti-
in¯ationary reputation inherited from its precursors, the national central
292 Otmar Issing
banks (NCBs). In addition, two pillars ± a reference value for the growth of a
broad money aggregate and a broad-based assessment of the outlook for
in¯ation ± are used to explain monetary policy decisions necessary to achieve
the price stability objective. The prominent role of money ± which we consider
as a reference value, not a target in the traditional sense intended in the
monetary literature ± underlines the principle of continuity and is rooted in
robust theoretical and empirical arguments accumulated over many decades
of research. Indeed, probably the only statement with which everybody would
agree is that in¯ation is, in the long run, a monetary phenomenon. However,
uncertainties in the stability of money demand would make it inappropriate to
automatically respond to deviations of the growth in money aggregates from
target. This is why the Council decided not to adopt a strict monetary targeting
framework. Instead, the prominent role of money as a reference value
is complemented by a broad-based assessment of the in¯ation outlook,
including both model-based forecasts and a variety of other relevant
indicators. Together these two pillars allow for a timely and ¯exible response
to a changing environment, while keeping the objective of price stability in
clear focus.
The strategy will help communicating with the public and organising policy
action in a coherent way. We do not see it as an invariable dogma. The ECB
closely follows the developments in economic research: this is a key factor in
our continuous assessment of the strategy itself. We look forward to a
continuous, productive exchange between the ECB and the academic world
on these issues.
Notes
1. I thank Frank Smets and BenoõÃt Mojon for their valuable assistance.
2. There remains uncertainty about the effect of bank dependence on the MTP. For
example Cecchetti (Chapter 5 in this volume) has emphasised that countries in
which ®rms are more bank-dependent will be more sensitive to the Eurosystem's
decisions to change interest rates through a kind of credit channel. However,
Schmidt has argued in chapter 6 in this volume that long-term relationship bank
lending, which occurs more naturally in a bank-dominated ®nancial system, has led
banks to insure customers against interest rate shocks.
3. See, for example, Calmfors (1998).
4. See, for example, Blanchard (1998) or Bean (1994).
5. In addition, Burda's Chapter 7 in this volume shows that there is not yet a consensus
as to how the institutional features of continental European labour markets
in¯uence macroeconomic outcomes.
References
Bean, C. (1994) `European Unemployment: A Survey', Journal of Economic Literature,
32(2), 573±619.
Blanchard, O. (1998) `European Unemployment. Shocks and Institutions', Baf® Lecture,
mimeo.
Concluding Remarks 293
Borio, C. E. V. (1995) `The Structure of Credit to the Non-government Sector and the
Transmission Mechanism of Monetary Policy: A Cross-country Comparison', in Bank
for International Settlements, Financial Structure and the Monetary Policy Transmission
Mechanism, CB 394, Basle.
Burda, M. (1999) `European Labour Markets and the Euro: How Much Flexibility Do We
Really Need?', Chapter 7 in this volume.
Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference Does EMU Make?', Oxford Review of Economic Policy, 14, 125±51.
Carlino G. and R. DeFina (1998) `Monetary Policy and the U.S. States and Regions: Some
Implications for European Monetary Union', Federal Reserve Bank of Philadelphia,
Working Paper, 98±17.
Cecchetti, S. (1999) `Legal Structure, Financial Structure and the Monetary Policy
Transmission Mechanism', Chapter 5 in this volume.
Kieler M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission?: A
Case not Closed', Economic Papers, 132, European Commission, Directorate-General
for Economic and Financial Affairs.
Maclennan, D., J. Muellbauer and M. Stephens, (1998) `Asymmetries in Housing and
Financial Markets Institutions and EMU', Oxford Review of Economic Policy, 14, 54±81.
Monticelli, C. and O. Tristani (1998) `What Does the Single Monetary Policy Do? A SVAR
Benchmark for the European Central Bank', DG Research of the European Central
Bank, mimeo.
Nickell, S. (1997) `Unemployment and Labour Market Rigidities: Europe versus North
America', Journal of Economic Perspectives, 11(3), 55±74.
Peersman G. and F. Smets (1999) `The Taylor Rule: A Useful Monetary Policy Benchmark
for the Euro Area?', International Finance, 2(1), 85±116.
Prati, A. and G. Shinasi (1997) `EMU and International Capital Markets', in P. Masson,
T. Krueger and B. Turtelboom (eds), EMU and the International Monetary System,
Washington, DC, International Monetary Fund.
Schmidt, R. `Differences between Financial Systems in European Countries: Conse-
quences for EMU', Chapter 6 in this volume.
Francesco Giavazzi
The wave of mergers and acquisitions that is sweeping the European banking
industry should not be a matter of indifference for those concerned with the
possibility that the monetary transmission mechanism may work asymme-
trically across EMU: for two reasons. First, banks, as is well known, are at the
centre of ®nancial markets in continental Europe, and thus of the transmis-
sion mechanism. Second, the consolidation of the European banking
industry, which is currently happening, may sharpen, rather than reduce,
the existing sources of crosscountry asymmetries.
Evidence of the importance of banks in the transmission mechanism inside
EMU is related to the extent to which monetary policy operates through a
`credit' channel, rather than simply through a `money' channel. As is well
known, when loans and bonds are imperfect substitutes in the balance sheets
of banks, following a squeeze in liquidity, banks reduce the amount of loans
they supply. Firms could turn to the bond market, but if bonds and loans are
imperfect substitutes the external ®nance premium will go up. This effect
[known as the `credit' channel (see, for example, Bernanke and Gertler, 1995)]
works on the supply side and ampli®es the more traditional demand effect of a
monetary tightening ± the change in interest rates which affects new marginal
spending by modifying borrowing conditions and by affecting asset prices,
and thus the market value of wealth.
Because of the large share of bank loans in the total debt liabilities of
European non-®nancial ®rms (85 per cent in Germany, 80 per cent in France
and Spain, 95 per cent in Italy, as opposed to 32 per cent in the United States)
the behaviour of banks is thus central to understanding the transmission
mechanism in Europe.
Reliable identi®cation of a `credit' channel requires the use of micro data:
macroeconomic time series are ill-suited to identify a `credit' channel from a
`money' channel in the transmission of monetary policy from the central bank
to banks. This is because the money channel works through banks' liabilities,
while the credit channel works through their assets, but assets and liabilities
are tightly related by accounting identities. For this reason, the evidence
proposed by macroeconomic studies which look at output and price
¯uctuations in response to shifts in the quantities of loans and deposits is
rarely decisive. On the contrary, microeconomic data allow one to ask whether
the responses of banks and ®rms to a shift in monetary policy differ depending
on their characteristics ± their size, for instance: small ®rms are more likely to
be liquidity-constrained and to depend on banks for ®nancing; similarly, small
banks ®nd it more dif®cult to insulate their loans' portfolio from a squeeze in
central bank liquidity, because a small bank typically cannot use bond
294
Concluding Remarks 295
holdings as a buffer (Kashyap and Stein, 1997; Kashyap, Stein and Wilcox,
1993). So far, the empirical evidence on the importance of bank characteristics
in determining the response of loans to a shift in monetary policy has been
limited to US data. Favero, Flabbi and Giavazzi (1999) have extended this
evidence to Europe.
Their ®ndings, albeit limited to a speci®c episode ± the EMU-wide monetary
contraction which occurred during 1992 ± point in two main directions. In
Germany there is clear evidence of a lending channel for all banks, except the
smallest ones ± that is, for over 80 per cent of the German banking sector. In
France, on the contrary, there is no evidence of a lending channel,
independently of the size of the banks considered. The large Italian banks
are similar to the corresponding German banks, in the sense that there is
evidence of a lending channel: the impact on the supply of loans of a change
in reserves is however twice as large as for the corresponding German banks.
The difference between Germany and Italy lies in medium-sized banks (whose
share of the domestic banking industry is larger in Italy than it is in Germany):
while there is evidence of a lending channel for medium-sized German banks,
this is not the case for the corresponding Italian banks. In Spain, there is
evidence of a lending channel for medium-sized banks: the impact on lending
of a cut in reserves is similar to that found in Germany.
The empirical evidence thus points to signi®cant cross-sectional and cross-
country differences in the response of individual banks to monetary policy ± at
least in the speci®c episode studied in Favero, Flabbi and Giavazzi (1999). In
Germany, Italy and Spain monetary policy operates via a lending channel in
an important segment of the market ± a segment that accounts for more than
80 per cent of the total German banking industry and about half of the Italian
and Spanish industry. The situation is quite different in France, where there is
no evidence of a lending channel.
Cross-country differences in the process of ®nancial intermediation could
be the result of varying national preferences and traditions. Consequently, a
consolidated, cross-border, ®nancial institution may wish to continue offering
different products in different markets. Similarly, the respective roles of
markets and intermediaries may be history-dependent in a way that will not
allow for rapid changes. Nevertheless, the creation of cross-border suppliers of
®nancial services, at a time when European consumers and ®rms are likely to
become more similar, would plausibly result in a homogenisation of ®nancial
practices across EMU.
The current consolidation of the European banking industry appears,
however, to be moving in the opposite direction. It is instructive to observe
what is happening in the light of the parallel experience in the United States.
Despite the massive consolidation which has occurred, concentration at the
local level has, if anything, decreased. Table C2.1 shows the Her®ndahl index of
the concentration of local markets for bank deposits in the United States:
consolidation has been accompanied by no signi®cant change in concentration.
296 Francesco Giavazzi
Notes: a The deposit Her®ndahl index is 10 000 times the sum of squared market shares based on
deposits of banks operating in MSA and non-MSA counties
Source: Berger, Demsetz and Strahan (1998), reproduced in Danthine et al. (1999).
Table C2.2 documents the characteristics of the consolidation which has so far
occurred in Europe. Unlike in the United States, most European banking deals
(half of all mergers or acquisitions in 1997) have involved institutions based in
the same country. Cross border activity has been limited to deals involving a
bank and a non-bank ®nancial institution (NBFI), mostly an investment bank,
an insurance company or an asset manager. While cross-border deals are
motivated by the search for experience in corporate ®nance and asset
management ± skills that are in scarce supply in continental European banks
± domestic deals are mostly driven by the search for size.
There are two reasons to be concerned. Competition is the ®rst. European
banks have a natural tendency to consolidate within national boundaries,
leading to industry concentration ratios much above those observed in the
United States. This is because of potential cost-cutting, culture and trust ± and,
indeed, the quest for market power at a time of insecurity and change: sheep
get closer together when in danger. In commercial banking, diversi®cation
gains explain the success of interstate consolidation in the United States. The
anaemia of the equivalent cross-border merger and acquisitions (M&A)
business in Europe is worrisome: while it can be explained by the fact that a
good deal of the gains from diversi®cation can be obtained within the borders
of individual European states, it also generates the concern that European
commercial banks will want to reach the higher minimum size in their
business simply by acquiring or merging with their national competitors. The
Table C2.2 Bank acquisitions in Europe, 1993±7 (value of all deals, US$ billion)
Total 19 40 122
Of which:
± Domestic bank/bank 9 24 60
± Cross-border bank/bank 1 8 7
± Bank/non-bank 9 8 55
References
Berger, A. N., R. S. Demsetz and P. E. Strahan (1998) `The Consolidation of the Financial
Services Industry: Causes, Consequences, and Implications for the Future', Board of
Governors of the Federal Reserve System, November, mimeo.
Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of
Monetary Transmission Mechanism', Journal of Economic Perspectives, 9, 27±48.
Danthine, J. P., F. Giavazzi, E. L. von Thadden and X. Vives (1999) `A Brave New World:
European Banking After EMU', London, CEPR.
Favero, C. A., L. Flabbi and F. Giavazzi (1999) `The Credit Channel and (A)Symmetries in
the Monetary Policy Transmission Mechanism in Europe: Evidence from Banks'
Balance Sheets', Milan, Boconi University, mimeo.
Kashyap A. K. and J. C. Stein (1997) `What do a Million Banks Have to Say about the
Transmission of Monetary Policy?, NBER Working Paper, 6056.
Kashyap A. K., J. C. Stein and D. W. Wilcox (1993) `Monetary Policy and Credit
Conditions: Evidence from the Composition of External Finance', American Economic
Review, 83, 78±98.
Claes Berg1
298
Concluding Remarks 299
When the central bank discusses how it intends to meet the price stability
target there are several different ways to do it. First, the central bank could
publish the in¯ation forecast for the target horizon, given an unchanged
monetary policy stance, and then adjust the interest rate accordingly. This is
done by in¯ation targeting central banks in New Zealand, the United Kingdom
and Sweden. The second and most common way is to regularly publish a
report which discusses price developments, indicators of economic activity
and in¯ation and in¯ation expectations, in a more general way, but not
publish a speci®c in¯ation forecast. This is done by many central banks around
the world, including the ECB.4 In my view, the in¯ation forecast is a crucial
input in monetary policy decisions and will always be very important for a
monetary policy directed at price stability. By publishing the forecast, the
central bank will increase the understanding of monetary policy. It will also
contribute to increased transparency and accountability, as the public at large
will be able to assess whether the stance of monetary policy is appropriate or
not.5
However, the ECB is a new central bank and I guess that it will take time and
experience to organise working routines. I expect that the Eurosystem will
develop its methods in order to be prepared to publish explicit forecasts some
time in the future.6
As was pointed out in Chapter 2 in this volume by Lars Svensson, the
formulation of the monetary policy strategy of the Eurosystem could be
clari®ed. Given the presentation of the reference value for M3 for 1999, it
seems that the implicit price norm is 1±2 per cent. It could be announced that
the ECB aims at a point estimate (for example, 1.5 per cent). In any case, a
symmetric in¯ation target, either a point target or a target range with equal
emphasis on the upper and lower bound, has some obvious advantages in
order to anchor expectations.
Another issue in relation to the ECB strategy ± which can be discussed ± is
the role of gradualism versus activism in interest rate setting. It is widely
accepted that policy makers facing uncertainty about the structure of the
economy should be more cautious when implementing policy. This is the
celebrated conservatism principle due to Brainard (1967). When there is
uncertainty about some structural parameters the policy response usually
becomes more cautious. In Svensson's Chapter 2, this no longer is always true.
When there is non-additive uncertainty and certainty equivalence no longer
holds, some covariance patterns for the parameters may make the optimal
response more sensitive when uncertainty increases. However, Svensson is not
very speci®c, in Chapter 2 at least, for which variables this may be the case in
practice.
One possibility that comes to mind is when there is uncertainty about the
persistence of in¯ation. Then, somewhat paradoxically, it may be optimal for
the central bank to respond more aggressively to shocks than if the parameters
were known with certainty. With persistence in in¯ation, a too small policy
Concluding Remarks 301
response this quarter will lead to deviations of the target not only next quarter
but also in forthcoming quarters. Therefore it may be optimal to pursue a more
aggressive monetary policy this quarter when the economy is hit with a shock,
in order to avoid bad outcomes in several periods (see So È derstro
È m, 1999).
Another possibility is when policy makers are uncertain about the state of
the economy and learn from the economy's reaction to policy. When the
private sector anticipates systematic attempts to incorporate this information
into policy, modest interest rate changes may prove ineffective. In a recession,
gradual policy initatives may elicit very little reaction. Because small interest
rate cuts are unlikely to end the recession, ®rms and consumers feel safe
waiting for rates to fall again before considering investing/consuming. A
vicious circle may develop in which the expectation that the policy could fail
leads investors/consumers to delay investment/consumption, thereby pro-
moting failure (see Caplin and Leahy, 1996).
This type of reasoning may have some bearing on the situation in part of the
Euro Area, for example in Germany, although nominal interest rates are very
low ± the IFO business expectations and business climate index has been
falling since the end of 1997.
Notes
1. I would like to thank Magnus Jonsson, Yngve Lindh, Peter Sellin and Lars Svensson
for their comments. The views expressed are solely the responsibility of the author
and should not be interpreted as re¯ecting the views of the Executive Board of
Sveriges Riksbank.
2. Regarding the second point, the RBC (or SDGE, stochastic dynamic general
equilibrium) literature have built sound theoretical models and confronted
simulated moments from these models with the data for many years. Even formal
statistical tests have been developed ± for example, SMM. McCallum's Chapter 1 in
this volume gives support for and follows this modelling tradition. Given this, one
question naturally arises (which is not discussed in Chapter 1): why is the central
bank not optimising? There seem to be two dominating strategies for modelling
monetary policy at the moment. On one hand, we have the optimising dynamic
general equilibrium approach where monetary policy usually only follows an (ad
hoc) monetary policy rule. On the other, we have a class of models where the central
bank is optimising but the agents' behaviour is not modelled in any detail. It would
be interesting to see a combination of these two modelling strategies. That is, models
where the central bank is optimising welfare, taking into account its effects on the
agents' behaviour, and at the same time, agents are optimising given the central
banks' behaviour. Of course, this is not an easy task since it involves modelling
dynamic games. But, nevertheless, it would most certainly give new and interesting
insights on how optimal monetary policy should be conducted.
3. A nominal anchor can be implemented in order to avoid both in¯ation and
de¯ation (see Berg and Jonung, 1999) for an account of the early Swedish
experience, 1931±7.
4. Another possibility would be to announce the monetary policy reaction function,
the policy rule or instrument rule, the central bank intends to follow. The rule
302 Claes Berg
should explain how the interest rate should react, given certain deviations of, for
example, actual in¯ation from target, deviations of actual output from potential
output and changes in real exchange rates. However, at least to my knowledge, no
central bank has announced such a reaction function. The reason, I guess, is that
there may be situations when monetary policy will have to deviate from the
announced reaction function.
5. An interesting task for future research is the following: how should the forecast be
aggregated over members when there is an executive board or a governing council?
Voting over forecasts may be very dif®cult. This problem has aspects related to the
problem ®rst discussed by Condorcet in the 18th century. In modern terms, it is
often referred to as the 'jury problem': to decide whether the accused is guilty or not
requires con¯ating the opinion of several experts, with varying competence, into a
single judgement. In practice, central banks ± be them in¯ation targeters or
monetary targeters ± solve this problem. But how is it done?
6. In the Swedish case, in¯ation forecasts have been gradually introduced (see Berg,
1999). The implementation and communication of monetary policy since 1993 can
be divided into three phases. In the ®rst phase, 1993±5, the in¯ation target strategy
was announced and established. However, during this period, in¯ation forecasts
were not published by the central bank; in the reports, the risks for future in¯ation
were stated in a more general way. In the second phase, 1996±7, an in¯ation-forecast
targeting was introduced. The Riksbank's own in¯ation forecasts were given more
weight in the communication of monetary policy. Forecasts for future in¯ation were
gradually introduced. In the third phase, from 1998, 'distribution forecast targeting'
(see Svensson's Chapter 2 in this volume) was introduced and explicit paths for
future in¯ation were published, surrounded by uncertainty intervals.
References
Berg, C. (1999) `In¯ation Forecast Targeting: the Swedish Experience', paper presented at
the seminar on 'In¯ation Targeting' in Rio de Janeiro, May, Sveriges Riksbank, Working
Paper.
Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience
1931±37,' Journal of Monetary Economics, 43 (3), 525±51.
Brainard, William (1967) `Uncertainty and the Effectiveness of Policy', American
Economic Review, 57, Papers and Proceedings, 411±425.
Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference Does the EMU Make?' Institute for International Economics Studies,
Stockholm University, Seminar Paper, 657.
Caplin, A. and J. Leahy (1996) `Monetary Policy as a Process of Search', American
Economic Review, 86 (4), 689±702.
Jonsson, M. and P. Klein (1997) `Tax Distortions in Sweden and the United States', in
M. Jonsson, Studies in Business Cycles (IIES, Stockholm University, Monograph Series,
No. 34).
È derstro
Ignazio Visco
1 Introduction
303
304 Ignazio Visco
will be structured around three themes. First, I shall give a brief summary of
current economic conditions and prospects in the Euro Area. Then I shall
sketch out what I see as some of the key issues and constraints which confront
the design of macroeconomic policy and monetary policy in particular and,
®nally, highlight some of the key questions concerning monetary policy in
the Euro Area.
The Euro Area experienced a slowdown in growth towards the end of 1998. On
a year-on-year basis GDP growth eased to 2.4 per cent in the ®nal quarter of
1998 (with almost no growth with respect to the third quarter) compared with
about 4 per cent in the ®rst quarter and a solid 3 per cent in the central part of
the year. Leading indicators suggest that activity remained sluggish in the ®rst
half of 1999. The slowdown in activity primarily re¯ected weaker external
demand, associated with recession in many emerging markets and Japan, and
low business investment. While still substantially higher than during the
trough of the 1993 recession, utilised capacity was declining and, while
consumer sentiments remain sustained, business con®dence was rapidly
deteriorating through the second half of 1998. However, weaker growth was
not common to all countries in the Euro Area. In 1998 France recorded its best
GDP growth performance since 1989 and periphery countries such as Ireland
and Finland continued to grow at a pace close to double that for the Euro Area
as a whole.
Divergent in¯ation rates within the Euro Area have also surfaced. In¯ation
measured by Eurostat, the Harmonised Index of Consumer Prices (HICP) is less
than 0.5 per cent in France, Germany and Austria and above 2 per cent in
Ireland and Portugal. The gap between the highest and lowest annual in¯ation
rate in January 1999 rose to 2.3 percentage points,1 compared with 1.4 percentage
points a year earlier. One might ask whether a disparity of this size might pose
a signi®cant problem for setting monetary policy. While some persistent
differences in in¯ation rates are plausible to the extent that they re¯ect a
catch-up process and short-term differentials could be signi®cant, owing to
differences in the relative cyclical position or tax changes, the appraisal would
certainly be more circumspect if the disparity in in¯ation were sustained as a
permanent difference.
Differences in regional price movements within long-established monetary
unions provide a gauge of the potential divergence of in¯ation rates in the
Euro Area. Large in¯ation differentials have seldom been observed in other
monetary unions and when they did happen, they were very short-lived
episodes. For instance, the differential between East and West Germany
following uni®cation was initially almost 10 per cent per annum, but fell
quickly to less than 1 per cent. In the United States, modest differentials of
Concluding Remarks 305
between 0.7 and 1.3 per cent per annum among state consumer price indices
have been apparent over the last thirty years. Likewise, the average differential
between the lowest and highest in¯ation rate in the major Australian cities
over ten-year periods is well below 1 percentage point; the same holds true
considering the differential across Spanish regions.
While in¯ation differentials in economically homogeneous areas are
unlikely to accumulate to large price level differences, in¯ation differentials
could persist between the more and less advanced Euro Area economies. As
is well known, productivity differentials between traded and non-traded
goods might not only generate sectoral in¯ation differentials, but differ-
ences in in¯ation rates between countries at different stages of develop-
ment. One might consider this to be a possibility even for the Euro Area:
indeed GDP per capita ranges between 12 000 Euros in Portugal and 29 700
Euros in Luxembourg (17 400 Euros for the area as a whole), even if the
variation is much less if one adjusts for PPP differences. Other indicators of
living standards ± such as the number of doctors, passenger cars, telephones
and television sets per 1000 inhabitants ± also vary to a non-negligible
extent.
While some evidence of divergent economic conditions and disparate levels
of development is apparent across the Euro Area, much progress towards
structural and cyclical convergence has been achieved. Indeed, among the
largest three economies in the Euro Area ± Germany, France and Italy ± which
account for some three-quarters of total output, the level of economic
development and living standards are very similar. The Maastricht criteria also
suggest that closer convergence has been obtained over more recent years.
Actually one of the most striking developments has been the rapid pace at
which Euro Area countries reduced their ®scal de®cits and accomplished the
nominal convergence criteria (in¯ation and interest rates) required by 1997
(the reference year). This achievement was con®rmed in 1998. Indeed, in
spring 1999 only Ireland, the Netherlands and Portugal (accounting for less
than 10 per cent of the Euro Area GDP) have in¯ation rates slightly higher
than 2 per cent. With Finland, they are the only Euro Area countries with a
positive and growing output gap.
In general terms, a monetary union is less susceptible to asymmetric
disturbances the more its regions are integrated with each other and
diversi®ed within themselves. Apart from the Maastricht convergence
criteria prior to the launch of the Euro, there are three key forces that
promote closer integration. These are the degree of trade interdependence,
the degree of intra-industry trade and closer income linkages, such as
increased foreign direct investment or closer ®nancial market interactions.
Over the past twenty years each has risen sharply and promoted closer
synchronisation of business cycles across Euro Area countries.2 This process
of integration is expected to continue in the years to come, and is possibly
accelerated by the introduction of the Euro.
306 Ignazio Visco
changing) with the launch of the Euro. Moreover, few historical series exist on
which to base fairly conclusive analysis. To give an example, how accurate is
the claim that real interest rates are near historical low levels in the Euro Area?
This appears to be the case when real rates are measured based on past
movements in the consumer price index (CPI), but quite a different picture
emerges when alternative de¯ators or in¯ation expectation measures are used.
For example, the short-term (long-term) real rate for the Euro Area using
headline HICP in¯ation is 2.3 (3.2) per cent. But, if a business investment
de¯ator were adopted, the measured real rate might be up to a full percentage
point higher (and even more, if producer prices were used). On the other hand,
excluding commodity and energy prices from headline in¯ation might lead to
measures of short-term real interest rates below 2 per cent.
In addition, it is well known that all CPIs have some inescapable positive
bias. In the United States estimates of the bias range between 0.5 and 1.0 per
cent per annum. The design and compilation of the HICP index, used by the
European Central Bank (ECB) to monitor in¯ation, attempts to limit the
source of known biases, but it is still believed to be positive. This implies that
the level of in¯ation at 0.8 per cent may in fact be closer to the lower end of the
range implicit in the ECB's interpretation of price stability, and this would
raise important questions concerning the stringency and symmetry of the
ECB's monetary policy objective.
While monetary policy in a single currency area, by de®nition, can respond
only to area-wide in¯ation pressures, sovereign ®scal policy can respond to
limit the contractionary or expansionary impacts of regional shocks. This
response is, however, limited by the provisions of the Stability and Growth
Pact. Based on a variety of techniques, further progress on budget consolida-
tion is required in order to avoid `excessive' de®cits and to provide adequate
scope for ®scal stabilisers to fully operate in the event of a slowdown in
economic activity. Most estimates suggest that a ®scal position on average
close to balance would be necessary to this end. This compares with structural
general government ®nancial balances that still exceed, for the average of the
Euro Area, 1.5 per cent of GDP.
Fiscal policy plans over the ensuing two to three years are targeting declines
in de®cits (from 2.3 per cent of GDP in 1998, to about 1 per cent of GDP). This
is in line with the principles of the Stability and Growth Pact. Furthermore,
most countries are still near or exceed the limit of the general government debt
to GDP ratio of 60 per cent speci®ed in the Maastricht Treaty. Although the
®scal plans embody tight controls on public spending, they are also based on
somewhat optimistic growth projections and in some cases `backloaded'. A key
risk to their realisation is a larger than expected and prolonged slowdown in
activity, which constrains automatic stabilisers from operating without
breaching the Pact's 3 per cent de®cit to GDP limit. Such a scenario, however,
would imply substantially lower GDP growth. OECD simulations suggest that
a 1 percentage-point reduction in output growth would result in an
308 Ignazio Visco
4 Conclusions
These remarks highlight the many uncertainties and much uncharted water
ahead concerning the framing of monetary policy in the Euro Area. To
complicate matters further we are still in a learning process about the
transmission channels and the effectiveness of monetary policy in the new
monetary area. Answers to, or at least a better and deeper understanding of,
the following questions might help reduce some of these uncertainties:
1. How effective would a small interest rate cut be in the current context? In
particular, what can be learned from US experience ± that is, what can be
said about the effects on economic activity stemming from con®dence,
expectations and asset price movements?
Concluding Remarks 309
With respect to the ®rst point, what would seem the most important is the
need to avoid possible cuts in interest rates ending up reducing rather than
improving con®dence. This calls for improved policy coordination efforts
which would also be helpful on the communication side: this would prevent
the perception that the easing of monetary policy would either re¯ect the fact
that the ECB had given up on the expectation that governments were ready to
act rapidly and effectively on the structural front, or signal a state of current
and prospective economic conditions much worse than expected.
The second point is, perhaps, even more dif®cult to deal with. This is simply
because there is still not much information on how the Euro Area economy
operates ± i.e. how economic decisions might differ under a common
monetary policy compared with the previous state of affairs. What might be
relevant here is probably the need to avoid the ECB's concerns about price
stability being perceived as biased towards less favourable monetary condi-
tions than would be needed in a situation of substantial current and
prospective slack in the Euro Area as a whole. To this end, the (perhaps
evolving) identi®cation of a lower bound for the de®nition of price stability
might be helpful (at least operationally). Moreover, it is important that the
ECB should not be perceived as feeling equally at ease with price changes
permanently situated at the bottom and at the centre of the range.
Finally, while many measurement problems must still be sorted out, both
with respect to price indices and monetary aggregates, a wide spectrum of
interest rates should be considered. In fact, one can always expect to ®nd
con¯icting signals concerning the stance of monetary conditions, depending
on the indicators used. In addition the Euro Area has virtually no historical
perspective against which benchmarks can be established. Consideration
should then also be ± and undoubtedly will be ± given not only to the levels
and spreads prevailing in the Euro Area member countries in the course of the
1990s, but also to those prevailing in other economies at different points of
the cycle, as well as over a longer time perspective.
Notes
1. Portugal has the highest in¯ation rate at 2.5 per cent and Luxembourg the lowest at
±1.4 per cent. The reported differential, however, excludes Luxembourg since the
rate in January has been in¯uenced by a one-off effect almost entirely owing to the
inclusion of the January discount sales in the price index for the ®rst time.
310 Ignazio Visco
2. The OECD has just published an in-depth assessment of the challenges facing the
European Single Currency Area, including an evaluation of the forces promoting
closer economic integration (OECD, 1999).
Reference
OECD (1999) EMU: Facts, Challenges and Policies, Paris, OECD.
Manfred J. M. Neumann
I will brie¯y take up two issues: (1) What type of ®nancial sector structure is
preferable? (2) Should the European Central Bank (ECB) be concerned about
differences in regional or national transmission mechanisms?
311
312 Manfred J. M. Neumann
role. This is highlighted by the following facts: (1) in Germany it takes 42 years,
on average, until a ®rm is admitted to the share market by a regulatory
committee that is dominated by banks; in the United States, in contrast, it takes
14 years on average; (2) in the United States market capitalisation, as a
percentage of GDP, is three times of what it is in Germany; (3) the share of bank
lending in total ®nance is much smaller in the United States than in Germany.
In line with these differences in ®nancial structure are the differences in legal
protection. In the United States the private investor in equity ®nance is better
protected by more extensive rights for controlling company management than
in Germany. As a corollary, creditors enjoy better protection under German
than under US law. How have the ®nancial cultures come to develop so
differently? A very probable explanation is the different role of universal banks.
Since the 1870s universal banks have dominated the German culture of ®nance
and this has in¯uenced German law while in the United States the banking
reform of the early 1930s put an end to the existence of universal banks and
promoted the fragmentation of the banking system.
But note that we lack systematic knowledge about the relative advantages as
regards the long-run impact of alternative structures of the ®nancing system.
All in all, it seems that bank-dominated Germany has not fared worse than the
more equity ®nance-oriented United States. While the long-term development
of economies obviously does not solely depend on the quality and structure of
the ®nancial system but on a host of other factors, too, it is interesting to note
that the German economy has not grown less, but faster than the Anglo±
Saxon economies. On average over the past ®fty years, real per capita GDP has
grown by 3.5 per cent in Germany but by about 2 percent in the United States
and in the United Kingdom.
regulation. Banks will become more similar across the Euro Area and move
into new areas of ®nance. As the Banking Report of the ECB (ECB 1999) shows,
banking intermediation is losing out to institutional investors, like investment
funds, insurance companies and pension funds. Thus, I conclude that the
transmission mechanism in the Euro Area will indeed become more similar to
that of the United States. We can therefore expect that monetary policy will
lose its strength to some degree in the longer run.
The Treaty of Maastricht assigns two objectives to monetary policy. The ®rst is
maintaining price stability. Provided that this objective is not endangered, the
European System of Central Banks (ESCB) is expected to support the general
economic policies in the Community. Though this second stipulation is open
to interpretation, it probably aims at business cycle stabilisation. The ECB puts
emphasis on the objective of maintaining price stability by calling it the
overriding objective.
With respect to this objective differences in regional or national transmis-
sion mechanisms are of no particular relevance. Consider that EMU starts from
a state of price stability and let us assume that the ECB has adopted a medium-
run strategy of supplying a stable monetary expansion path ± relative to the
evolution of money demand. Under those equilibrium conditions all member
countries are affected to the same degree. Asymmetries in the transmission
mechanisms will not matter, given that the ECB's forecast errors as regards the
appropriate medium-run path are likely to be comparatively small. But note
that we cannot be sure that the ECB has adopted a medium-run strategy. The
ECB has announced that it will take into consideration its reference value for
the growth of the money stock M3 and at the same time a host of other
economic indicators.
In contrast, differences in the transmission mechanism will matter if the
ECB produces monetary shocks which by de®nition are unanticipated. This
will become particularly relevant should the ECB try to ®ght a business
downswing or upswing. The regional economies will be hit differently,
implying that their relative positions as regards competitiveness in the single
market will be affected. This applies especially to small and medium-sized
®rms who traditionally rely on bank ®nance. Moreover, it seems that those
member economies where the bank lending channel is strongest will be hit the
harder if their business cycles run counter to the average cycle showing up in
the EU-11-wide aggregates. This will be a problem for the smaller members of
EMU whose data enter the aggregate data with small weight.
The question is: what lesson should the ECB draw from this perspective? It
goes without saying that national business cycles are none of its business.
Monetary policy has to be based on the EU-11 aggregates and nothing else.
314 Manfred J. M. Neumann
Reference
ECB (1999) Possible Effect of EMU on the EU Banking System in the Medium to Long Term,
Frankfurt/Main.
Name Index
Note: page numbers in bold refer to main contributions in this book; page
references to ®gures and tables are given in italics.
Index compiled by Sue Lightfoot.
315
316 Name Index
Tobin, J. 211
142, 143, 144, 154±5
Tristani, O. 289
Weber, A. A. 4, 131±56, 161,
164±8
Uhlig, H. 2, 51±9
White, W. R. 172
Uribe, M. 115
Wieland, V. 67, 72, 94 n14, 115
Williams, J. C. 72
Vickers, J. 78
Ä als, J. 3, 107±11
Vin
Wolman, A. L. 94 n14
Visco, I. 8, 303±10
Woodford, M. 13, 46, 80, 132, 168
Wyplosz, C. 151
Vlaar, P. J. G. 183
Walsh, C. E. 27
Yellen, J. L. 259
Note: page references to ®gures and tables are given in italics, e.g. 20f, 32t.
aggregate demand and supply 113, 124 channels of monetary policy 210±11
asset price channel see channel of channel of relative prices 211±12
relative prices credit channel see credit channel
exchange rate channel 211
Bank of Canada 64, 77 interest rate channel 211, 212, 248±9
Bank of England collective bargaining 264±5, 267, 277,
forecast targeting 76, 78±9, 82, 86, 92, 280
106 Consumer Price Index 66±7
in¯ation targeting 66, 67, 86, 95±6 consumption 125±7, 126±7f
n37, 96 n38, 105, 106 credit channel 125, 204±6, 212±15,
bank reserves 174±5 228±9, 247±8, 249, 294
banking industry balance sheet channel 213, 214, 228
change 206, 290±1 bank lending channel 176, 213±14,
consolidation 294, 295±7, 296t, 312±13 228, 295
cross-country differences 171±2, 180t, cross-border competition 205
214, 295 France 249, 295
France 291, 295 Germany 227±8, 228f, 241±2, 243f,
Germany 219, 220, 222, 295, 312 247, 249, 295
(see also Deutsche Bundesbank) Italy 295
health 178, 179, 179t Spain 295
Italy 295 strength 178, 181t, 186, 187, 196,
lending channel 176, 213±14, 228, 197, 198±200, 199f
295 United Kingdom 227±8, 228f, 241,
liability diversi®cation 204±5 243f, 247, 249
size and concentration 177±8, 177t,
181t, 196±7, 197t, 198
de¯ation 68, 115±24
Spain 295
Deutsche Bundesbank 61, 106, 135,
United Kingdom 219, 222 (see also
146, 153
Bank of England)
United States 295, 296t, 312 (see also
Federal Reserve System (USA)) ECB see European System of Central
see also Bank of Canada; credit channel; Banks
European System of Central EMU see European Monetary Union
Banks (ESCB); New Zealand, ESCB see European System of Central
Reserve Bank of; Sveriges Banks
Riksbank European Monetary Institute 67, 73
European Monetary Union (EMU) 195,
capital mobility 254, 254t 210, 303, 306
channel of relative prices 211±12 economic development 305
Index compiled by Sue Lightfoot.
319
320 Subject Index
growth 304
250
integration 305
225±9, 228f, 241, 243f
249±50
214±15, 285
104±5
see also banking industry
82
forecast targeting see targeting
M3 de®nition 205
banking industry 291, 295
290
credit channel 249, 295
313
Germany
313±14
transparency 92
Bundesbank)
history 116±24
241±2, 243±5f, 246, 246t, 248
311±12
effect 124, 127
complementarity and
222, 241
259±61, 260t, 261t, 278, 304±5
161±3, 165±6
labour unions 264±5, 267, 277, 280
148f, 149f
and impact of monetary policy 173,
France 248
structures
Germany 227±8, 228f, 241±2, 243±5f,
lending channel see credit channel
246, 246t, 248
liquidity traps 113±15
United Kingdom 227±8, 228f, 242,
policy
diagnostics 31±6, 32±3f, 32t, 34f, 35f,
interest rates
36f, 54±5
European links
maximum likelihood 47
167
second moments 47
empirical evidence 144±8, 145t, communication 62, 92, 108, 109, 111
monetary policy)
208±9
smoothing rules 135±8, 137t, 138t,
Euro Area 306±9
142, 146, 154
impact on output and prices 173t,
wealth effect 124±5
181±3, 182f, 184t, 187t, 198±200,
policy
in¯ation targeting 61, 63±6, 67±8, 88,
investment 53±4
intermediate targeting 61, 86±8, 103,
Italy 295
104, 110
labour markets
legal systems and 173, 173t, 186±8,
280±2
monetary targeting 61